Skip to main content

Full text of "Studies in the linguistic sciences"

See other formats

If too tightly bound to photocopy, 
please take to the Circulation Desk so 
personnel can remove plastic posts. 


1?!!!^,^' Return or renew all Library Materialsl The Minimum Fee for 
each Lost Book is $50.00. 

The person charging this material is responsible for 
Its return to the Hbrary from which it was withdrawn 
on or before the Latest Date stamped below 
Theft, rnutilation, and underlining of books are reasons for discipli- 
nary action and may result in dismissal from the University 
To renew call Telephone Center, 333-8400 


AUG 2 5 199:: 

AUG 2 7 19SI 

^UG 2 ,? 1994 
AUG G 9 m 

««mws list cwL* 

APR 2 

; ^ u 


APR 2 9 


JAM 10 2*^^^^^/ 


.SEP 1 20^3 

LI61— O-1096 

Studies in 


s2 FALL 1984 COPY 2 



EYAMBA G. BOKAMBA French colonial language policy in Africa 

and its legacies (Pari I) 1 

GEORGE N. CLEMENTS Binding domains in Kikuyu 37 

CHET A. CREIDER Language differences in strategies for the 

interactional management of conversation 57 

NICHOLAS FARACLAS Rivers Pidgin English: tone, stress, or 

pitch-accent language? ^^ 

HUSSEIN ALI OBEIDAT Relative clauses in Standard Arabic 

revisited ' ' 

HERBERT STAHLKE Derivational conditions on morpheme 

structure in Ewe ^ " ' 

/■v ij i j U i tyoO 

ALEKSANDRA STEINBERGS Loanword incorporation processes: 

exampley^r^<^l r^^/l^'^er IL-LfNOI* "^ 

BRENT VINE African 'shadow vowels': a descriptive survey 127 

JENNIFER J. YANCO Modifiers in Bantu: evidence from Spoken 

Lingala ' ■'" 

ELUZAI M. YOKWE Arabicization and language policy in the 

Sudan 149 

Department of Lingui^ics 
University of Illinais J^ 





EDITORS: Charles W. Kisseberth, Braj B. Kachru, Jerry L. Morgan 

REVIEW EDITORS: Chin-W. Kim and Ladislav Zgusta 

EDITORIAL BOARD: Eyamba G. Bokamba, Chin-chuan Cheng, Peter 
Cole, Alice Davison, Georgia M. Green, Hans Henrich Hock, Yamuna 
Kachru, Henry Kahane, Michael J. Kenstowicz and Howard Maclay. 

AIM: SLS is intended as a forum for the presentation of the latest original 
research by the faculty and especially students of the Department of 
Linguistics, University of Illinois, Urbana-Champaign. Especially invited 
papers by scholars not associated with the University of Illinois will also be in- 

SPECIAL ISSUES: Since its inception SLS has devoted one issue each year to 
restricted, specialized topics. A complete list of such special issues is given on 
the back cover. The following special issue is under preparation: Linguistic 
Studies in Memory of Theodore M. Lightner, edited by Michael J. Kenstowicz 
and Charles W. Kisseberth. 

BOOKS FOR REVIEW: Review copies of books (in duplicate) may be sent to 
the Review Editors, Studies in the Linguistic Sciences, Department of 
Linguistics, University of Illinois, Urbana, Illinois, 61801. 

SUBSCRIPTION: There will be two issues during the academic year. Requests 
for subscriptions should be addressed to SLS Subscriptions, Department of 
Linguistics, 4088 Foreign Languages Building, 707 S. Mathews, University of 
Illinois, Urbana, Illinois 61801. 

Price: $5.00 (per issue) 



The person charg-ing this material is re- 
sponsible for its return to the library from 
which it was withdraw'n on or before the 
Latest Date stamped below. 

Theft, mutilation, and underlining of books are reasons 
for disciplinary action and may result in dismissal from 
the University. 
To renew call Telephone Center, 333-8400 





L161— O-1096 

This special issue of 

studies in the Linguistic Sciences 

is dedicated to 


Molakisi monene mpe mokambisi wa lokumu 



Eyamba G. Bokamba: French colonial language policy in Africa and its 

legacies (Part I) 1 

George N. Clements: Binding domains in Kikuyu 37 

Chet A. Creider: Language differences in strategies for the inter- 
actional management of conversation 57 

Nicholas Faraclas: Rivers pidgin English: tone, stress, or pitch- 
accent language? 67 

Hussein Ali Obeidat: Relative clauses in Standard Arabic revisited. . 77 

Herbert Stahlke: Derivational conditions on morpheme structure in Ewe 97 

Aleksandra Steinbergs: Loanword incorporation processes: examples 

from Tshiluba 115 

Brent Vine: African 'shadow vowels': a descriptive survey 127 

Jennifer J. Yanco: Modifiers in Bantu: evidence from Spoken Lingala. 139 

Eluzai M. Yokwe: Arabicization and language policy in the Sudan . . . 149 


Since the appearance of the first issue of Studies in the Linguistic 
Sciences devoted to African Linguistics (Vol. 6, No. 2, 1976) edited by 
Charles W. Kisseberth and myself, the Program in African Linguistics in 
the Department of Linguistics at the University of Illinois has experienced 
a significant growth in all areas. In 1976, for example, the Department 
had only a handful of students specializing in African Linguistics; today 
there are over twelve graduate students specializing in different areas of 
linguistics: phonology, sociol inguistics, syntax, and historical linguis- 
tics. This increased interest is a reflection of and parallels the expanded 
research interest of the faculty. While in 1976 there were only two faculty 
members with research interests on African languages (Bokamba and Kisseberth), 
today there are five: E.G. Bokamba (Bantu syntax, sociol inguistics, and 
general African linguistics), C.C. Cheng (Phonology/tonolog)^ , B.B. Kachru 
(sociol inguistics), M.J. Kenstowicz (phonology), and C.W. Kisseberth (phono- 
logy/tonology, general African linguistics). Visiting Africanist faculty 
members provide additional strength to the program from time to time. 

This expansion in the regular faculty research interest has made it 
possible for the Department to offer a diversified program in African lin- 
guistics and thereby to attract a variety of graduate students. Since 1976, 
for example, the Department has granted over ten doctorates to specialists 
in African linguistics, most of whom are teaching at universities throughout 
the world. In 1979 the Department, in collaboration with the African Studies 
Center, hosted the 10th Annual Conference on African Linguistics to mark both 
tfte tenth anniversary of the establishment of this professional gathering and 
that of the program in African Linguistics. 

The collection of papers included in this issue is an example of the 
continuing research interest that has developed in the last eight years or so. 
The papers represent a broad spectrum in the field: four papers on phonology 
(Faraclas, Stahlke, Steinbergs, Vine), three on syntax (Clements, Obeidat, 
Yanco), and three on sociol inguistics (Bokamba, Creider, Yokwe). Five of the 
ten papers are by current and former faculty (Bokamba and Stahlke) and students 
(Obeidat, Steinbergs, Yokwe) of this Department. The collection includes three 
papers selected from the 10th Annual Conference on African Linguistics (Clements, 
Stahlke, Vine). More papers from that conference could have been included, but 
we were not able to contact several authors in time for this publication. 

The program in African Linguistics at Illinois began with the teaching 
of Swahili by Chin-W. Kim in 1968. But the progress described above would not 
have been possible without Braj B. Kachru's vision and unwavering support 
throughout the period of his tenure of office as head of the Department of 
Linguistics (1969-79). It was decided at the 10th Annual Confrence on African 
Linguistics in 1979 that the proceedings of the conference would be dedicated 
to him. Although this collection does not include many of the papers presented 
then, we would still like to think of it as representing the "spirit" of the 
proceedings of that conference. It is with great pleasure that we dedicate 
this issue to Braj as an expression of our appreciation for his scholarship, 
teaching, and leadership. 

Eyamba G. Bokamba 
Urbana, IL 
May, 1985 

Studies In the Linguistic Sciences 
Volume 14, Number 2, Fall 198A 

Eyamba G . Bokamba 

This paper discusses the Impact of the French colonial lan- 
guage policy on education and its legacies in the former French 
colonies, with particular emphasis on Western Africa. It shows 
that French language policy was culturally, politically, and 
economically motivated. The paper argues that in spite of the 
well-articulated colonialist objectives of this policy, inde- 
pendent Francophone African states have maintained its status 
quo. As a result, little effort has been made to modernize 
and promote indigenous African languages to serve national 
functions. The lack of African-based language policies vis-a-vis 
education has contributed to poor academic outputs and continu- 
ing high illiteracy rates in the region. To remedy this situa- 
tion, it is suggested that Francophone African states reevaluate 
their inherited language policy and adopt new policies that will 
respond better to national developmental objectives. 

1.0 Introduction 

Much of the published research on language policy and language planning 
(hereafter LP & LP) in Africa in the past two decades has focused on the 
historical development aspects and on the necessity for adopting African 
language-based policies that are consistent with nationalism, educational 
objectives, and socio-economic development (cf Bokamba and Tlou 1977, Ansre 
1976, 1978, Andrzejewski 1980b)l. Little attention, however, has been paid 
to the examination of the question of the Impact of African language policies 
on formal education, literacy, and language development (cf. Bokamba 1981, 
198A) . A discussion of this question is crucial to our understanding of 
recent developments in education and language policies in the continent 
particularly in countries such as Ethiopia, Somalia, Tanzania, Zaire, 
Madagascar, Nigeria, Senegal. Furthermore the articulation of comprehensive 
and objective language policies for the future progress and educational 
planning of Africa is vital. The Insight gained from such an examination 
would undoubtedly contribute to a better formulation of LP and LP theory. 

1.1 Objectives and scope of study . The purpose of this study is to 
examine the Impact of the French colonial language policy on education and 
its legacies in Francophone regions of Africa, particularly Western Africa. 
The paper surveys briefly the evolution of the French colonial language 
policy from 1826 to 1959, and examines its effect on three specific areas of 
language and education in the region: formal education, literacy education, 
and African language development or modernization. 

In the course of the discussion it is shown that Francophone Western 
(i.e., former French colonies in Central, West, and Northern) Africa has 

negatively distinguished itself in three ways: First, it has the highest 
average Illiteracy rate of any comparable region in the continent. Second, 
its school drop-out and class-repeater rates are the highest in the conti- 
nent (cf. Unesco 1977, Barnes 1982). And third, its lingua franca are among 
the least developed in Africa. 

Drawing on comparative data from other regions in the continent, an 
attempt is made to ascertain the causes of these problems. It is concluded 
that while rampant illiteracy and academic inefficiency are common phenomena 
in Africa, the particular difficulties in education and language development 
experienced by Francophone Western Africa are largely attributable to the 
French colonial language policy and its legacies In education. The paper 
concludes with a discussion of the implications of these findings for LP and 
LP. We begin with an overview of the literature on French colonial language 

1.2 Overview of the literature . Since the beginning of the 1960s 
there have been a number of studies devoted in part or entirely to French 
colonial language policy in Africa. Some of these studies have described 
the French language policy as being motivated by socio-political considera- 
tions whereby Africans were subjected to a policy of cultural assimilation 
through language and education for the purpose of rapid power consolidation 
(cf. Crowder 1967, Spencer 1971, Awonlyi 1976, Sylla 1978). Others have 
characterized it as an instance of linguistic Imperialism and ethnocentrlsm 
(Calvet 19^4) ; while still others have viewed it as the result of perceived 
mutual socio-econom.ic interests between the colonizers and colonized 
(Alexandre 1963, Turcotte 1981). 

More specifically, Alexandre (1963: 53-54) states that the imposition 
of French as the sole language of administration and education in colonial 
Francophone Africa was simply an extension of French monolingual policy 
adopted in the 16th century; and that its continuation to the detriment of 
African languages in the 1950s was due In part to the expressed wishes of 
African intellectuals to use French as the medium of instruction in a 
French-based system. 2 

Over-all, the picture that emerges from the literature Is that French 
colonial language policy In Africa was dictated by (a) politico-cultural 
considerations; (b) African intellectuals' desire to benefit from the saiie 
educational system as that offered in France; and (c) various local lin- 
guistic factors. The first factor has generally been interpreted in the 
literature as being la force motrioe (cf . Spencer 1971, Calvet 197A, 
Bokamba and Tlou 1977) . How accurate is this view? Were there other 
factors that motivated French language policy In Africa? To answer these 
questions we need to review the evolution of the French colonial language 
policy, giving particular emphasis on the key documents and socio-political 
factors that shaped it. 

2.0 French Colonial Language Policy 

The French colonial empire extended through much of Northern and 
Western Africa, and comprised the twenty countries listed in Table 1 below. 
These nations, most of which acceded to political independence at the begin- 
ning of the 1960s, had an estimated total population of about 65,075,000 in 


Table 1: French Colonies and Protectorates in Africa* 


Independence date 

Fstimated 1^60 

Affars & Issas (now Djibouti) 


Central African Republic 


Dahomey (now Benin) 

French Cameroon 

French Congo (People's Rep. of 

French Guinea 
French Sudan (Mali) 

Ivory Coast 
Upper Volta (now Burkina Faso) 

27 June, 1977 
3 July, 1962 

13 August, 1960 

11 August, 1960 
1 August, 1960 

1 October, lti61 

15 August, 1960 

2 October, 1958 
22 September, 1960 
17 August, 1960 

7 August, I960 

26 June, 1960 

28 November, 1960 

12 March, 1968 

2 March, 1956 

3 August, 1960 
20 August, 1960 

27 April, 1960 
20 March, 1960 

5 August, 1960 





*Source (of population estimates): UNESCO Statistical Yearbook, 1964 

The French colonial language policy in these countries evolved through a 
series of decrees, ordinances, official memoranda/communiques, and practical 
decisions made by colonial administrators in the colonies. 

2.1. Implementation of the language policy . The most important piece 
of legislation that shaped the evolution of the French colonial language 
policy in Africa was the metropolitan ordinance of Villers-Cotteret which 
was issued in 1539 by King Francois I. This ordinance made French the ex- 
clusive official language of the French Kingdom, thus disallowing the use of 
Breton, Basque, Flemisch, Occita, German, Provencal, etc., in official func- 
tions (Alexandre 1963: 53). In 1634 Cardinal Richelieu established the 
French Academy to regulate the language (Alexandre 1963: 54), thus enhancing 
its authority and prestige. 

The ordinance of Villers-Cotteret was extended to all French colonies 
when France became a major colonial power in the 19th century. The evidence 
for the evolution of this policy in Africa has been ably compiled by Denis 
Turcotte (1981b) in his Repertoire Chronologique de la Politique Linguistl- 
que en Afrique Francophone . According to Turcotte, the first documented 
evidence of the extension of the ordinance of Villers-Cotteret to Franco- 
phone Africa appeared in 1826. On July 14 of that year, decree No. 14 
authorizing the establishment of a girls' school In Saint-Louis, Senegal, 
stipulated in its article 8 that (Turcotte 1981b: 28): 

(1) La langue francaise sera seule employee par les eleves. 

On October 18, 18A8, decree No. 95 of the director of administrative 
services in the colonies referring to local elections stated in its article 
9 that because of the insufficient number of Arabic interpreters to be 
provided for the election prlcincts, all the election materials would have 
to be written in French (cf. Turcotte 1981b: 31). 

The success of the Koranic schools, which were taught in Arabic, and 
the schools' unwillingness to teach French as a subject became a matter of 
concern to the colonial authorities. To curb the Moslem community's influ- 
ence and force it to comply with the colonial language policy, ordinance 
No. 96 regulating "Moslem schools" was issued on June 22, 1857. Article 5 
required all teachers to take or send all pupils aged 12 and above to eve- 
ning French classes at the government or Catholic church schools (Turcotte 
1981b: 31): 

(2) Les maltres d'ecole musulmanes seront obliges de condulre ou 
d'envoyer, tous les jours, a la classe du solr (soit celle de I'ecole 
laique, soit celle des f reres) tous les eleves de 12 ans et au dessus. 

After ordinance No. 96 failed to achieve the desired results, the 
French colonial administration issued a stronger and more comprehensive 
ordinance on February 28, 1870, consisting of eleven articles. This ordi- 
nance made knowledge of French not only a requlrment for pupils and teachers 
in Moslem schools, but also a prerequisite for the establishment and con- 
tinuation of such schools. Students and teachers were required to demon- 
strate proficiency in French after a period of two years; failure to do so 
led to the dismissal of the student from the Koranic schools and to his 
subsequent enrollment in a government or Catholic mission school (Turcotte 
1981b: 37-38). Schools which failed to provide instruction in French after 
the grace period lost their licence to operate. Monetary and non-monetary 
Incentives were given to schools whose students received the best results on 
the yearly competitive examination in French. 

In short, the primary objective of ordinance No. 85 of February 28, 
1870^ was to force the Koranic schools, which were until 1857 the only ones 
allowed to teach in a language other than French, to serve as an agent for 
the spread of French in colonial Francophone Africa. The ultimate goal of 
the colonial administration was to completerly absorb the Koranic schools 
into the French educational system and thereby eliminate them. These 
schools could not have been outlawed directly, because such an action would 
have been construed as a violation of the philosophical tenets of the 
French Revolution. 

The most Interesting part of ordinance No. 85 is the preamble where the 
French colonial administration's politics of assimilation emerges clearly. 
We quote this passage in its entirety, along with the first article, because 
of its implications on future LP and LP activities in the region (Turcotte 
1981b: 37-38): 

(3) Nous, Gouvernaur du Senegal et dependences, 

Vue I'arrete du 22 juln 1857 sur les ^coles musulmanes; 

Conslderant que le but de 1 'administration de la colonie, 
en regularlsant par ledlt arret^ 1' institution des ^coles ^ 
musulmanes, a ete de aheroh^r a s'ssimiley les enfants indigenes; 

Que ce but n'a pas ete atteint lusqu'lci, par suite de 
1 'indifference apportee par les maitres d'ecoles; 

Attendu que le moyen le plus efficace d'arriver a ae vesultat 
parait ttve d'exigev desormats que aes maitres hahituent les 
enfants d' aomprendre et a parler la langue fvanpaise; 

Attendu qu'il convient, dans ce but d'astreindre a certaines 
conditions les Indlvidus qui demanderont a tenlr des ecoles 
arabes ; 

Sur la proposition du directeur de I'interieur, 

Le consell d 'administration entendu, 

Avons arrete et arretons: 

Article ler. Nul ne pourra, a I'avenir, obtenir 1 'autorisation 
de tenir une ecole musulmane si, en conformite de la prescription 
de 1 'arret! du 22 juin 1857 precitee, il n'habite Saint-Louis 
depuis sept annees, et s'il ne produit un certificat de bonne 
vie et moeurs du maire de la ville, et ne justifie savoir parler 
le frangais devant un jury d 'exarren compose 

du chef du 2e bureau de la direction de I'interieur, 

du maire de la ville, 

et du cadi, chef de la religion musulmane. 

L ' autorisation sera retiree si le titulaire en devient indigne. 
("Emphasis added.) 

After the imposition of this ordinance on the Moslem schools, the community pro- 
tested in vain only to be told firmly by the colonial administration that 
the exceptional status they enjoyed previously with regard to the language 
policy was actually "an abuse" of the law, and must be terminated immediate- 
ly. The Moslem community leaders recognized that ordinance 85 and its pre- 
decessor were aimed at curbing not only their influence in education, but 
also at coercing them to become agents of the spread of French and its cult- 
ure. The ultimate goal of the French adinistration was to force the Islamic 
community out of general education, and thereby confine it to religious 

Although ordinance No. 85 was specifically issued to regulate the oper- 
ation of Koranic schools, its broader intent was to secure complete control 
of general education for the colonial administration. This exercise in 
power consolidation was motivated by three important factors in French colo- 
nial politics. First, the French government in the 1800s espoused the view 
that the education of the people, whether colonized or free citizens, was 
the responsibility of the government, not of any private organization (cf. 
Bokamba 1984). Second, the French considered the church to be the "enemy of 
reason, the stronghold of conservatism and an obstacle to development and 
progress" (Awonlyi 1976: 32). In view of this, the French colonial adminis- 
tration in Africa could not entrust the education of the African children 
with any religious organization nor allow any competition against the state 
in this area. 

Third, and most importantly for our purpose here, the French colonial 
administration in Africa viewed education as the most effective tool for 
colonizing the Africans, and made a serious though selective effort to ex- 
tend it to the target population (Bolibaugh 1972: 5). The Governor-General 
of the French West African colonies (i.e., I'Afrique Oacidentale Fran^aise, 

A.O,F,)j Mr. E. Chaudle, stated this point eloquently and forcefully in his 
circular letter of June 22, 1897, addressed to his territorial administra- 
tors and conunanders. He stated in paragraphs four and five of this letter 
that (Turcotte 1981b: 51): 

(4) L'eoole est, en effet, le moyen d 'action Ze plus sup qu'une nation 
civilisatrice ait d'acqueinr a" ses idees les populations enaove 
primitives et de les elever graduellement jusqu'a elle. L'eoole 
est^ en un mot, I'element de progres par excellence. C est aussi 
I 'element de propagande de la cause et de la langue fran^aises le 
plus certain dont le Gouvemment puisse disposer. Ce ne sont pas, 
en effet, les vieillards imbus des prejuges anciens, ce ne pas 
meme les hommes faits, plies deja a d'autres coutumes, que nous 
pouvons esperer convertir a nos principes de morale, S nos regies 
de droit, a nos usages nationaux. PoiO' accomplir avec succes 
aette oeuvre de trans formation y c'est aux jeunes qu'il faut s'adre- 
sser, c'est t 'esprit de la jeunesse qu'il faut penetrer et c'est 
par I'ecole seule, que nous y arriverons. 

C'est vous dire, Monsieur 1 'Administrateur , quelle importance 
j 'attache au developpement de I ' instruction publique, a" la diffu- 
sion de la langue et des idees franaaises et au bon fonationnement 
des ecoles. (Emphasis added.) 

Governor Chaudie's statement is an eloquent illustration of the French assim- 
ilationist philosophy which became a dominant influence in French Africa.^ 
It is evident from the above passages that education for the colonized people 
was not an end in itself, but rather a means through which acculturation and 
servitude were to be achieved. In this regard, the French language was seen 
not only as the vehicle par excellence for the spread of French culture, but 
also as the ultimate beneficiary of thousands, and eventually millions, of 
new speakers. 

We shall not elaborate on this point here. Suffice it to say that a se- 
ries of executive decrees, reinforced by regular official memoranda, were 
issued to promote the use of French in all administrative and educational 
functions, on the one hand, and to proscribe the use of African languages in 
such functions, on the other hand, throughout the colonial period (Stumpf 
1979, Turcotte 1981b). These regulations were further backed up by various 
monetary and non-monetary rewards to schools, students and functionaries who 
demonstrated proficiency in French and its culture. No such inducements 
were provided for proficiency in African languages, except that regional 
administrators were encouraged to familiarize themselves with the major lan- 
guages of their subjects for administrative efficiency. 

The use of African languages in education and administration was active- 
ly discouraged by the French colonial administration, because African lan- 
guages were viewed as an obstacle to the objectives of cultural assimilation 
in French. Specifically, these languages were seen as perpetuating racial 
and cultural differences between the colonized and colonizers. It is for 
this reason that Governor-General Chaudie's statement cited in (A) above 
emphasized the transformation of the African youths through education, rather 
than that of the old people ("vieillards") whom he considered to be "imbued 
with old prejudices" and "bent to other customs" and values. French colonial 

administration wanted to erase these differences through its policy of 
assimilation. African languages had, therefore, to be prohibited from play- 
ing any instrumental role in education and administration. 

Governor-General W. Ponty, Chaudie's successor, pursued the same stra- 
tegy of using education as an agent of the spread of French and its culture. 
In a lengthy circular No. 82c, dated August 30, 1910, Governor Ponty insist- 
ed on the importance of the policy of assimilation (cf. Turcotte 1981b: 73), 
and emphasized the wisdom of pursuing it through formal education fTurcotte 
1981b: 74): 

(5) L'eaole est le meilleux' instrument du progves; a 'est elle qui sert 
le mieux les interims de la cause fran^aise et qui en transformant 
peu a pen la mentalite de nos su^.ets nous permettra de les acque- 
rir a nos idees sans heurtev auoune de leurs traditions. Nul n' 
ignore en effet que 1' etude du francais est le remede le plus 
efficace qui puisse etre oppose au fanatisme et 1' experience nous 
apprend que les musulmans qui connaissent notre langue sent mains 
inbus de pre^uges que leurs coreligionnaires qui ne sax>ent que 
I 'arabe . 

Governor Ponty instructed his administrators to multiply the number of state 
schools in all regions, especially in the rural communities, so as to pre- 
empt and/or reduce the expansion of the Koranic schools. 

2.2 Language and assimilation policy . The French colonial administra- 
tion's policy of assimilation raises questions of motives. I'Jhy did the 
French government assume exclusive responsibility for the education of its 
African subjects, instead of sharing it with or delegating it to private and 
religious organizations as other major colonial powers (i.e., Belgium, 
Britain, and Portugal) in the continent did? Why did the French colonial 
administration espouse the policy of assimilation, instead of an evolution- 
ary or laissez faive one as practiced by their Belgian and British counter- 

The French colonial politics and policies in the region were guided by 
three major factors: cultural imperialism or ethnocentricism, economic con- 
siderations, and military resources (Morgenthau 1964, Calvet 1974, Turcotte 
1981a, b, Lokulutu 1982). The third factor, which is not immediately relevant 
to this study and will not be discussed further, was particularly applicable in 

the period from World War I to World War II (1918-45). 

The French, it is commonly acknowledged in the political literature, 
believed that they had a civilizing mission to carry out in Africa (cf. 
Morgentahu 1964, Crowder 1967, Lokulutu 1982). They felt that their civil- 
ization was superior to that of the Africans, and that the best way to bring 
them to par was through an active policy of political and cultural assimil- 
ation. Education a la fvanqaise, initially aimed at the sons of chiefs and 
subsequently extended to a highly selected youth population, was the medium 
through which the policy was Implemented, as evidenced in the passages in 
(4) and (5) above. Viewed from the humanitarian principles of egalite, 
fratemite, et justice for all irrespective of race and creed advocated by 
the French Revolution, the assimilation policy V7as seen as the best approach 
to erase the cultural and attitudinal differences that existed between the 

colonized and colonizers so as to develop a common culture — the French way 
of life (Mumford 1935, Crowder 1967). The ultimate goal, as observed ear- 
lier, was to facilitate the submission of the African people to French colo- 
nization - both culturely and politically. 

Whether the policy of assimilation was generally considered by the 
French government to be a success, remains to be determined. The evidence 
that has emerged in the past twenty-four years, i.e., since the advent of 
political independence of most of the states in the region, suggests string- 
ly, however, that assimilation was partly a failure and partly a success. 
It was a failure to the extent that it did not transform the former colonies 
into an overseas France and its people into overseas Frenchmen. The differ- 
ences between the former colonized people and the colonizers remained large- 
ly unchanged. As Morgenthau (1964: xxi) aptly observes with regard to West 

(6) When the obvious ties of colonialism were broken one goal of the 
first generation of French-speaking West African party leaders was 
achieved. But only one. Independence changed the political 
rulers. It did not wipe out the differences — economic, cultural 
and political — between the rulers and the ruled. After independ- 
ence, to meet the expectations of the mass of the people and 
provide occasions for further social mobility, economic develop- 
ment became the order of the day. It was the goal of the genera- 
tion which aspired to the succession. (Emphasis added.) 

Economic development and independence have proven to be difficult goals to 
reach for most of the former French colonies in Africa. One factor that ac- 
counts for this situation is that the economic structures of the colonies 
were integrated into, or rather, made dependent upon, the French (metropo- 
litan) economic structure (Ake 1982, Lokulutu 1982). This brings us to the 
second factor that motivated French colonial policies in Western Africa. 

While cultural imperialism and military considerations were undeniably 
important in the conduct of French colonial politics, the most important 
factor which guided not only the conduct of that politics but also motivated 
the colonization of the region was economic. Governor-General W. Ponty, in 
his letter of August 30, 1910, cited in part in (5) above, notes the import- 
ance of the French educational policy to this aspect of colonization. He 
states (of. Turcotte 1981b: 74): 

(7) Heme si on n' envisage que le point de vue commercial de notre colo- 
nisation, il faut reconnaitre que I'instruction sert les interets 
de la Metropole plutot que de leur nuire. Ainsl que je I'ai dit, 

I 'instruction en transformant le gout de nos sujets augmente aussi 
leurs appetits, c'est-a-dire leur puissance de consomation et les 
oblige a travailler. En creant des ecoles nous contribuerons 
done a I ' accroissement de la rechesse dans le pays et nous obtien- 
drons des indigenes une collaboration d'autant plus active que 
mieux renseignes sur nos intentions a leur egcvcd (et) ils auront 
une oonfiance plus marquee dans notre autorite et deviendront 
mains dociles aux suggestions interessees des marabouts ignorants 
ou fanatiques. 

II me parait d'ailleurs inutile d'insister encore une fois sur les 
avantages de tout ordre que nous pouvons avoir c^ reoruter sur 
place les fonctionnaires necessaires a notre Administration ou les 
ouvriers indispensables au developpement de notre outillage eaonomi- 
que. (Emphasis added.) 

It is evident from these statements that French colonial policy of 
assimilation through education was motivated by ultimate economic concerns. 
The twenty African countries identified in Table 1 above represented a size- 
able market not only for French manufactured goods, but also a significant 
source of cheap raw materials and cheap labor force. As long as this was 
the case, the logic of their educational and language policy is perfectly 
defensible: They had specific economic objectives and devised the most 
appropriate policy to achieve them. 

Considered from the point of view of economic integration whereby a 
dependency relationship was created between the so-called metropole and the 
colonies, the policy of assimilation was a great success. As in other colo- 
nies in the continent, French colonies provided cheap labor and the raw 
materials (e.g., agricultural, mineral and oil) that made the French economy 
prosper. The economic infrastructure developed in the colonies was oriented 
towards the export of raw materials to France and other Western European 
countries. As in most other African colonies, virtually no industries were 
set up in the French colonies prior to the end of World War II, and the few 
that were established between 1945 and 1959 were mainly extractive. Manu- 
facturing industries were located, instead, in France. 

As a result of this economic structural relationship, France and the 
world market dictated not only what to produce in virtually all sectors, but 
also determined the prices to be paid for the commodities exported (Ake 
1982). This economic 'assimilation' worked so well during the colonial 
period that the leaders of pre-independence Francophone Africa saw no need 
to change it. In fact, they approved, with the exception of Guinea (former- 
ly French Sudan)," the formation of a French economic community and a common 
monetary zone under the French franc. On February 9, 1959, the eve of poli- 
tical independence for most of the twenty countries, this relationship was 
sealed by the adoption of French as the sole official language of the French 
Community (Turcotte 1981b: 131): 

(8) Article unique. La langue officielle de la Cormw.naute (franaaise) 
est la langue franpaise. 

With a few exceptions (see sections 2.3 and 3.0 below), French has remained 
the sole official language of administration and education in Francophone 
Africa today. 

2.3 Language Policy Legacies and their Raison d'etre . As has been 
shown, the use of African languages in the education of Africans was pro- 
scribed initially by the extension of the ordinance of Villers-Cotteret of 
1539, and subsequently by a series of colonial executive decrees. Up to 
1910 the decrees prohibiting the use of African languages were largely 
implicit in their mention of these languages, but this situation changed 
from 1911 onwards when several such decrees were issued to enforce the es- 
tablished language policy. The first decree of this type appears to be 


No. 1207 issued on October 1911, in connection with the founding of materni- 
ty schools in Senegal. Article 4, line 4 stipulated (Turcotte 1981b: 80): 

(9) L'emploi des idiomes locaux est rigoureusement evite. 

The enforcement of this policy was so strict that the printing or publica- 
tion of books in African languages for use in the so-called "village 
schools" was subjected to a 12.8% tax (Stumpf 1979: 82). Imported books in 
other languages (e.g., German, English) were levied the same amount of tax 
by the custom service. Mission schools were allowed to use indigenous lan- 
guages, but their educational functions were viewed by the French as strictly 
religious. The ordinance of February 14, 1922, spelled this out (Spencer 
1971: 543): 

(10) General education must be carried in French The Coranic 

schools and catechist schools are authorized to provide exclusive- 
ly a religious education in the vernaculars. Such schools are not 
considered as institutions of public education. 

The continuation of this language policy after World War II raised 
questions in the minds of some liberal French parliamentarians, especially 
in light of the establishment of the French Union or Community, consisting 
of France and its colonies, and the de Qvcce recognition of the equality of 
all the people in the Community. Following the Brazzaville Conference of 
January 1944, the French Constitution of October 27, 1946, which governed 
the new Community, recognized the equality of the cultures of the member 
countries and advocated the establishment of a democratic system within the 
Community (cf. Morgenthau 1964, Lokulutu 1982). This spirit of equality 
and democracy is evidenced in the preamble to the Constitution (Lokulutu 
1982: 276): 

(11) France forms with the people of overseas a union established on 

the equality of duties and rights, without racial or religious 
distinction. The French Union is composed of nations and peoples 
who put in common and coordinate their respective civilizations 
to increase their well-being and guarantee their safety. Faithful 
to her traditional mission, France intends to lead the peoples she 
has taken in charge to self -administration and democratic manage- 
ment of their affairs. Setting aside all systems of colonization 
founded on arbitrary powers, France guarantees to all equal access 
to public functions and to individual and collective exercise of 
the rights and liberties proclaimed herein. (Emphasis added.) 

Whether this declaration had any substance remains a debatable question. 
Evidence from other parts of the 1946 Constitution and France's behavior 
from this period onwards indicate clearly that she had no intention of treat- 
ing other members of the Community as equal partners. She planned to remain 
not only the colonial power, but also the dominant political, cultural, and 
economic power in the Community. The French education system was to serve 
as the model for the educational systems of the colonies (both in Africa 
and elsewhere) , and the language policy established before the Brazzaville 
conference was to remain in force throughout the colonial period. Access to 
public functions or positions, especially in higher administrative posts, 
and job upward mobility were determined by one's education and competency in 


French: the more one's education mirrored that provided in France, the 
greater his chances for obtaining good employment and a commensurate salary. 
In a word, one's degree of success as an educated person depended on the ex- 
tent to which he or she assimilated to the French education and culture 
(Morgenthau 1964, Crowder 1967). 

Alexandre (1963: 54) reports on an event that occurred in the French 
parliament at this time which he claims contributed to the retention of 
French as the sole medium of instruction in the African colonies. According 
to this source, some French nationals brought up for discussion the question 
of the use of African languages in African education, with a view to chang- 
ing the existing language policy. The change was strongly resisted by the 
African representatives in the parliament on the grounds that it would 
create two unequal systems of education, one for Africans and the other for 
French, and thereby deny the former an equal opportunity to a French, eduaa- 
tion (Alexandre 1963: 54). Moreover, they viewed the attempt to introduce 
the use of African languages as media of education as a ploy to keep their 
children backward. Alexandre (1963: 54) goes on to observe that not only 
did the African representatives, except for Senghor, demand the extension 
of the use of French throughout the schools in the countries concerned, but 
also that language courses in Ewe offered under U.N. pressure in Togo were 
taken only by foreigners. An attempt by "African intellectuals" in Paris 
in the 1950s to find a solution to the question of the proper place for 
African languages and French in African education met with all sorts of ob- 
stacles. As a result, Alexandre (1963: 55-59) concludes, the status quo 
was maintained. 

In view of the parameters of success discussed above, it is not a all 
surprising that African representatives in the French parliament and many 
other intellectuals in Paris construed French education, with French as the 
exclusive language of instruction, as the best preparation for personal and 
national development. As long as France remained the colonial power of 
Francophone Africa and thereby dictated the conditions for success, it would 
have been foolhardy to adopt a language policy that would have affected the 
education provided to the colonized people. 

It is surprising that most Francophone African nations, except for 
Algeria, Morroco, and Tunisia where Arabic is a co-official language and 
Madagascar where Malagasy serves as the national language, have maintained 
the colonial language policy after almost twenty-five years of political in- 
dependence. The main difference between the French colonial language policy 
and the current language policies is that instead of advocating assimilation 
as the primary motivation, the avoidance of "tribal conflicts" or uprisings 
and access to world development through a language of wider communication 
(LWC) are advanced as arguments against the use of Afracan languages in 
education (cf. Bokamba and Tlou 1977). Arguments such as the following in 
favor of the retention of French and the exclusion of African languages are 
commonly heard in Francophone African states (cf. Turcotte 1981a: 65-68): 

(12) En tout etat de cause, le choix du franfais comme langue de 

travail, done comme langue officielle, laisse toute latitude a 
chaque Etat concern! d'utiliser cumulativement sa langue nation- 
ale. Je dots toute fois a la vevite de dire qu'en ce qui conceme 
mon pays (la Cote d'lvoire) , I 'adoption du franaais, par I'artiole 


premier de notre Constitution, a sans doute ete I 'un des facteurs 
d'unite qui ont favorise I'aboutissement heureux et si rapide de 
I'oeuvre de oonstruotion nationale dont Son Excellence le Presi- 
dent Felix Houphouet-Boigny avait fait un des premiers themes de 
son action. Le francais, librement accepte par nous, a ete un 
faateur de cohesion. . .a I'interieur de la cSte d'lvoire ou il a 
favorise le regroupement de nos quelque cent et^nies. .. (Emphasis 
added . ) 

This statement, according to Turcotte (1981a: 66), was made on April 26, 
19 76, by Mr Philippe Yace, then president of the Ivory Coast's National 
Assembly. In November of the same year, in an interview published by the 
newspaper Fraternite-Hebdo , Mr Jules Nea, then Minister of Cultural Affairs 
in the Ivory Coast national government, was quoted as having argued against 
the teaching of any Ivory Coast languages in the schools (Turcotte 1981a: 
66-67). He maintained that there is no single Ivory Coast language that 
could be chosen at that time to serve this purpose, and that the task of 
choosing 4 or 5 languages to be eventually introduced as subjects in the 
schools had been given to a research team at the University of Abidjan. In 
the meantime, Mr. Nea continued, French will have to remain the official 
language for national development considerations (Turcotte 1981a: 67): 

(13) Le probleme (de I ' enseignement d'une langue ivoirienne) est tres 
aomplexe et sa complexite n'a pas echappe a la Commission Nation- 
ale de la Reforme de I 'Enseignement. Il est peu probable, dans 
un delai plus ou moins rapproche, qu'une langue nationale puisse 
etre introduite dans les tcoles: la multiplicite des groupes 
ethniques, la diversite de nos dialectes ne favorisent pas I ' 
adoption d'une langue ivoirienne unique. On pourrait songer a 
choisir quatre ou cinq langues representatives de chaque ensemble 
de groupes ethniques. Ce travail est confie aux chercheurs de 1' 
Universite. A I'heure actuelle, le dioula vernaculaire et le 
baoule-agni sont timidement enseignes a 1 'Universite Nationale. 
La prudence est indispensable dans un domaine aussi delicat et 
complexe — qui va lentement, va surement dit-on. Mais il ne faut 
pas oublier que la Cote d'lvoire a choisi un developpement ouvert 
sur le monde exterieur: la necessite d'utiliser une langue inter- 
nationate s'impose par de telles considerations. Le franpais est 
non seulement la langue de I'economie, de I ' administration mais 
aussi de la plupart de nos ecrivains. (Emphasis added.) 

It is precisely this type of attitude that has permitted the perpetua- 
tion of the French colonial language policy in Senegal, Ivory Coast, Algeria, 
Morocco, Tunisia, Gabon, Congo-Brazzaville, and other former French colonies 
in Western Africa. While there are undeniable short-term advantages in the 
retention of French as the exclusive official language and medium of instruc- 
tion in most of the countries, the long term disadvantages entailed by the 
policy far outweigh the present benefits. Let us consider some of these 
problems, with particular reference to education and African language devel- 

3.0 Language Policy, Education and Language Development 

Access to education in the French African colonies was highly restricted: 


only a small percentage of the eligible population was selected for ad- 
mission into the available elementary and secondary schools. Of these, an 
even smaller percentage managed to complete their elementary and secondary 
education (Bolibaugh 1972). Facts cited by this author in connection with 
his study of education in Guinea, Mali, Senegal, and Ivory Coast are very 
revealing with respect to the inefficiency and restrictive nature of the 
French educational system. According to Bolibaugh (1972: 17), 

(lA) In 1957, the numbers (sic) of students, including Europeans, who 
received their baccalaureat were as follow: Guinea — 5, Ivory 
Coast — 69, and Senegal — 172. In 1960, 31 Malians were awarded 
the baccalaureat. 

Since there were virtually no university or other post-secondary training 
institutions in the colonies during the colonial era, a few qualified second- 
ary school graduates were sent to France to continue their studies. A con- 
sequence of this educational infrastructure was that most of the countries 
found themselves with a mere handful of secondary and post-secondary gradu- 
ates to shoulder the responsibilities of national administration at the ad- 
vent of their political independence in the early 1960s. 

Since that time. Francophone African states, like their counterparts 
elsewhere in the continent, have come to regard education as the most appro- 
priate vehicle towards national development. As a result, education has be- 
come the single most expensive item in the national budgets of most govern- 
ments, absorbing between 15 to 37 percent of such budgets as of 1973. Table 
2 below provides an indication of the growth of the educational budgets of 
selected nations in the region from 1960 to 1973, and shows what percentages 
of these budgets are devoted to teachers' salaries by each country. 

Table 2: 

Public Expenditure on Education in Selected Francophone 
Nations in Western Africa 


Total expenditure as % of 
Gross National Product Budget 
1960 1965 1973 1960 1965 

Teachers' salaries as % 
of public expenditure 
1973 1960 1965 1973 




Congo (P. Rep.) 



Ivory Coast 



















6 73.5 







Source : Adapted from UNESCO 1976: Conference of Ministers of Education of 
Member States. Education in Africa since I960: A statistical 


The second and third most expensive budgetary items in these countries, as 
elsewhere, are administration and defense, respectively. It will be noticed 
from Table 2 that the largest portion of the educational budget in each 
country was spent on teachers' salaries. This means that other areas of 
education, e.g., class-room facilities, libraries, students' scholarships, 
housing, and staff development received very little support. 

The wisdom of this type of allocation of scarce resources is highly 
questionable, especially for any developing country. The ultimate question, 
however, is how cost-effective have Francophone African nations been in their 
education? In other words, how productive and efficient has the educational 
system been under the present French-based language policy, given the high 
level of expenditure made by the nations concerned? Further, what impact, 
if any, has the present system had on literacy and African language develop- 
ment? As indicated at the beginning of this study, these questions have 
rarely been addressed in the literature. The discussion that follows is 
largely based on my own research and draws on circumstantial evidence found 
in UNESCO and other recent publications by individual scholars on education 
in Africa. 

3.1 Language policy and general education . Language policy in Franco- 
phone Western Africa has affected general education in at least four differ- 
ent ways since the introduction of Western education in the region: (1) ad- 
mission and promotion criteria; (2) learning strategies; (3) extension or 
application of knowledge; and (4) academic performance. Let us consider each 
of these aspects of the problem briefly. 

It is a commonly known fact that French education is highly elitist 
rather then mass-oriented: only the best qualified students get admitted and 
are maintained in school. In transplanting this philosophy of education to 
its African colonies and eventually getting them to maintain it after the ad- 
vent of political independence, France circumscribed educational development 
in Francophone Africa. While initial admission in primary one is open to all 
eligible children on the basis of space availability, promotion from one 
grade to another is strictly dependent on the pupil's performance in French 
and other subjects which are all taught in French. A consequence of this 
system of education is the disproportionately high wastage rates found in 
Francophone Africa. 

Wastage is shown in the number or percentage of pupils that (a) repeat 
classes because of failure in one or more subjects, or because of failure in 
an admission examination, and (b) drop out of school because of poor perform- 
ance at some level or lack of space in the next level of education. Table 3 
cited by Barnes (1982: 8) from UNESCO (1980) and slightly adapted here for 
the purpose of this study, shows the magnitude of the problem as evaluated 
on a yearly basis from 1967 to 19 77. 

As evidenced in Table 3, the percentages of class repeaters in Franco- 
phone states are consistently higher than those found in Anglophone states. 
When repetition rates are disaggregated to show a grade per grade develop- 
ment not only is this Francophone vs. Anglophone contrast maintained sharply, 
but it also becomes evident that the first and last grades of primary educa- 
tion are the most critical weeding stages in the cycle. Specifically, a 
higher percentage of the age cohort repeat the first and last grades. These 


Table 3: Total percentage of repeaters in primary education in selected 
African Nations (UNESCO 1985). 












French Speaking 


































Cent. African Rep. 













































Ivory Coast 

















































































































Upper Volta 





















English Speaking 





























































































Table 4: Total repetition rates in percentage 
latest year available (UNESCO 1983). 

by grade in primary education. 



Total # of 


e Rer 

eaters by 










French Speaking 






















Burund i 










Cent. African Rep 




























































Ivory Coast 














































































































Upper Volta 




















English Speaking 









































Sierra Leone 


















































facts are shown in Table 4 above (UNESCO 1983: III, 131-35). 

The causes underlying these facts are all related to the question of lan- 
guage of instruction. First, unlike in Anglophone countries where selected 
indigenous (African) languages are used as media of instruction in primary 
grades one through three, with English being offered as a subject from grade 
one to six or seven and then serving as language of instruction from grade 
three or four, in Francophone countries French serves this function from 
grade one onward exclusively. In other words, it is simultaneously taught as 
a subject and used as the language of instruction. Such a practice would be 
natural and acceptable if French were the language of the target population 
from which the pupils come. This is not the case in any former French colony, 
however. Instead, French is used mainly as the language of work, not of 
daily communication in the community, and its use in other functions requir- 
ing an LWC is restricted to a small percentage of the population of any of 
these nations. Further, French is learned almost exclusively in school. 

In Senegal, one of the most Francophone countries in the continent be- 
sides Algeria, French is spoken only by 11% of the population, while Wolof 
is spoken and understood by 80%, according to a statistical survey reported 
in Dumont (1983: 25ff ) . Except for Algeria, whose settler population repre- 
sents a significant portion of the French-speaking community, it is doubtful 
that any other nation can be shown to have a higher percentage of French- 
speaking population. 8 In the absence of language statistical surveys, it is 
impossible to provide a more accurate assessment of the popularity of French. 
As with other official languages in the continent, however, the extent to 
which French is spoken can be gauged indirectly on the basis of statistics 
on literacy and educational achievements. In Francophone Africa, as will be 
seen later, illiteracy is very high (ca. 80%) and educational achievement 
are low. 

To appreciated the negative impact of French-based education in Africa, 
one must understand the context in which the prospective student lives. The 
typical elementary and secondary school student in Western Francophone Africa 
lives in a multilingual community where either the mother or father, and 
often both, are illiterate and non-conversant in French. Even if one of the 
parents is, or both are literate, French is rarely the language of communica- 
tion in the family. Instead, one or two African languages are used as medium 
of intra- and inter- family communication. Often one of these languages is 
a mother tongue, and the others may be lingua f rancas . French in such sit- 
uations is often the third or fourth or fifth language that the child encoun- 
ters in his/her community, and one that (s)he can successfully avoid using 
until (s)he has to communicate in a classroom situation. In summary, French 
is characteristically a foreign language whose special functions in the soci- 
ety make it a remote medium of communication for the child's or student s 
daily activities. 

Because of this socio-linguistic situation, the child who enters primary 
school without prior knowledge of French is forced to become and will contin- 
ue to be a parrot for much of the primary school. His/her learning strate- 
gies are often reduced to simple memorization without understanding — a behav- 
ior that has also been ascribed to many pupils in English-speaking Africa 
(cf., e.g., Afolayan 1976). The pupil's natural learning abilities are fur- 
ther handicapped by his/her incompetence in French composition, thus making 


it difficult for him/her not only to take good notes, but also to understand 
written work. Consequently, the pupil cannot relate his/her classroom-ac- 
quired knowledge to the daily life in his/her community, and cannot perform 
well academically. If we recall that it takes a child between four to five 
years to master his/her native language (that is constantly spoken at home 
and in the community) ," it is not surprising that the performance of the 
African child in French in a largely non-French speaking population often 
turns out to be poor relative to the number of years spent on studying the 

The pupil or student's difficulties are compounded by poorly qualified 
teachers who, in many instances, have a mediocre mastery not only of the sub- 
jects they are required to teach, but also of the language of instruction 
itself. This situation obtains both in primary and secondary education 
throughout much of the continent (cf. Bokamba 1976, Bokamba and Tlou 1977, 
Thompson 1981). As a result, drop-out rates continue to be very high in 
both primary and secondary schools. Table 5 documents this phenomenon on 
selected Anglophone and Francophone African states at the primary school 
level (UNESCO 1980, cited in Barnes 1982: 9). 

Table 5: 


Total percentage of cohorts starting primary education around 
1965/66 and around 1976/77 reaching the final grade of the cycle, 
ranked in descending order according to survival for the last 

1st 2nd No. of 
Cohort Cohort Grades 


Ivory Coast 
















Upper Volta 
































19 76 





















19 76 









19 75 

















10 20 30 40 50 60 70 80 90 lOO'h 


This Table should be interpreted as follows. The complete length of 
each bar shows the percentage of the age cohort that would have reached the 
Ath grade under normal circumstances, while the blank (i.e., white) portion 
of the bar indicates the percentage that actually reached the last grade of 
the primary education cycle (Barnes 1982: 8). It will be noticed from this 
table that the Francophone countries, with the exception of Senegal, have 
the lowest survival rates. Such drop-out rates are disastrous not only for 
the educational system in general, but also for the achievement of literacy 
for which formal education remains the primary vehicle. 

The drop-out rates between the elementary and secondary education 
cycles, which are not documented systematically anyvhere to our knowledge, 
are considerably higher. The end of primary education represents not only 
the end of schooling for the vast majority of the children, but also the 
most wasteful stage in the educational systems in most Francophone African 
nations. It is the most wasteful stage because the children who fail the 
secondary school admission examination are successful primary school gradu- 
ates. They are generally prevented from attending general secondary school 
because they failed to pass one of the two admission examination questions: 
French or mathematics . 

Attrition rates within the secondary education cycle for the few sur- 
vivals continue to be so high that the percentage of those who reach the 
final year becomes insignificantly small, as Table 6 on the next demonstrates. 
This is a common problem throughout the continent. But what it suggests for 
our purposes here is that the imported system of education and language of 
instruction are unproductive and exceedingly costly for Africa. Given the 
socio-linguistic conditions under which the child learns and uses the lan- 
guage of instructions and the elitist-based philosophy of education adopted 
from France, it is unlikely that Francophone Western Africa will see signifi- 
cant changes in the academic performance of their children in the near fu- 
ture. Changes in the language policy vis-a-vis education, however, will 
positively affect the pupil's performance, as the facts from the United 
Republic of Tanzania show. 

3.2 Language policy and literacy development . Language policy in 
Francophone Africa has had a considerable negative impact on the continuation 
and development of illiteracy in the region. Generally speaking, literacy 
education has never been a priority area in much of Africa, except in Tanza- 
nia, Somalia, and Ethiopia for the last ten years or so. Most African states 
have chosen, consciously or unconsciously, to use general/ formal education 
as the best means to achieve literacy. As a result, very little attention 
has been given to literacy programs: most states spend less than 0.5% of the 
education budget on them and provide them very little other resources (e.g., 
personnel and teaching materials) to make them viable (Omolewa 1981, Bray 
1981, Bokamba 1984). Because of the high attrition rates experienced by 
African educational institutions, few countries succeed in raising their lit- 
eracy rates through this avenue. 

In Francophone Western Africa where school wastage rates are higher than 
in most other regions, illiteracy is equally rampant: 65-90% as of 1980 (of. 
Bokamba 1984: 21). The major factor accounting for this situation is the 
language of instruction. Unlike in Anglophone and other Francophone coun- 
tries (e.g., Zaire, Rwanda, Burundi) where literacy education is conducted 


Table 6: Percentage dls 

tribution of 

secondary school enrollment by grades 

in Selected African states 















































Central African Republic 





























































Ivory Coast 




















































































Sierra Leone 

















































United Rep of Cameroon 









United Rep of Tanzania 








Upper Volta 
































Source: UNESCO (1983: III-204-203) 


first in indigenous languages and then in the official language, literacy 
programs in most Francophone states in Western Africa are provided in French 
at all phases. Guinea, Mali, Tunisia, Algeria, Morocco, and to a limited ex- 
tent Senegal, are the only countries known to us to be using African lan- 
guages in literacy and other programs of adult education. In Senegal it is 
only Wolof, one of the six selected "national languages," that is currently 
being used on an experimental basis: a final decision is awaiting a govern- 
mental decree (cf. Botti, et al. 1978, Dumont 1983). 

As in the case of formal education, the results of the few established 
literacy programs in the region have remained meager. The situation has 
been worsened by the primary school drop-outs who relapse into illiteracy 
after a period of non-use of French. Thus when the progress made against 
the battle on illiteracy in the region is measured against other regions in 
the continent. Francophone Western Africa emerges as the least progressive, 
as Table 7 shows (Bokamba 198A: 21). 

Table 7: Growth in literacy in selected African states 


1979 1981 

Adult Literacy % circa 
1960 1976 1980 









*Central African Republic 
















*Ivory Coast 































Sierra Leone 






















































































































*Indicates Francophone Western Africa state. 


While the role played by non-language factors In the development of literacy 
can be shown to be significant, the language of instruction remains the most 
influential factor in this regard. 

3.3 Non-development of indigenous languages . The use of French as the 
language of instruction in literacy programs in most states in the region 
follows in part from the language policy they have adopted, and in part from 
what may be termed a language development problem. The exclusive use of 
French as the language of administration and education in the former French 
colonies caused the study of the African languages in the region to be neg- 
lected both before and after the accession to political independence. Afri- 
can languages were seen by the French colonial administration not only as 
instrumentally useless, but also as harmful to the objectives of assimilation. 
Consequently, French linguists never studied them seriously so as to provide 
grammatical descriptions for at least the major languages as was done in 
British and Belgian Africa. The few available grammars or grammatical 
sketches on the languages of the region, except for Arabic, were primarily 
the work of missionary-linguists (cf. Cole 1971). 

Although the perception of the usefulness of African languages in the 
region has changed and is continuing to change slowly, their over-all instru- 
mental value has remained relatively the same as during the colonial period. 
The teaching of African languages in elementary and secondary schools is 
viewed by both students and parents as a waste of time, because the languages 
do not have a market value. That is, they do not enhance the chances of the 
learner to secure employment (Turcotte 1981a, Dumont 1983). Further, the 
promotion of any African language to serve as a subject of instruction or a 
national language for the purpose of radio and television services is viewed 
by politicians as a divisive undertaking. Recall in this respect Mr. Nea's 
statement cited in (12) above. 

Each Francophone country in Western Africa has recognized either honor- 
ifically or in practice between two to six "national languages." These in- 
clude Agni-Baule, Diola, Senufo, Bete, and Yakuba in Ivory Coast; Wolof, 
Serer, Fula, Diola, Mandingo, and Sarakole in Senegal; and Bambara, Malinke, 
Diola, Fulani, Songhai and Tamashek in Mali (B. Dumont 1973, Turcotte 1981, 
P. Dumont 1983). Further northwest, viz. in Mauritania, Morocco, Algeria, 
and Tunisia, Arabic and Berber are the dominant lingua francas. In spite of 
the dominant role that these "national languages" have played as the princi- 
pal media of communication among the populations, their study has been and 
continues to be neglected. In a language resource survey undertaken by the 
African Studies Center at the University of Illinois' library in 1983, it was 
found that only one-fifth (7/35) of the languages for which the library had 
holdings were spoken in Francophone Western Africa. The remaining 28 lan- 
guages were spoken mostly in the former British and Belgian Africa, as can 
be seen from Table 8 on the next page. 

While this survey is admittedly non-exhaustive, it does, nonetheless, 
provide a reasonable estimate of the level of development attained in the 
description of and publication in African languages. The availability of 
dictionaries, grammars, and readers in a given language is one of the best 
indications of the extent of its development. If we assume that there are 
on the average three national languages spoken in each Francophone nation 


Table 3: Publications In African Languages at the University of Illinois, 1983 
Language Main location Dictionaries Grammars Readers 



South Africa 


















Northern Africa 












North West Africa 















Ghana, Togo 



































Zaire, Congo 






Zaire, Congo 










Mandoka/Bambara Mali 






Sierra Leone 
























Uganda , Rwanda 























Eastern Africa 






Zambia, Zimbabwe 












South Africa 





Senegal, Gambia 






South Africa 


















South Africa 




(1) Languages for which more than 50 titles are accessioned; (2) languages for 
which holdings are less than 5 items In each category. 


in Western Africa, the publication of reference resources in only seven of 
then respresents very little progress, especially when compared to other 
regions in the continent. In fact, if Duala and Ewe, which are also spoken 
in Anglophone states in the region, are omitted from consideration, the total 
number of languages with dictionaries, grammars and readers identified in 
Table 8 falls to five. 

The reasons for this paucity of publications are not difficult to deter- 
mine. As indicated earlier, French colonial administrations discouraged and 
neglected the study of African languages: they had no use for them and 
therefore saw no need to describe them. Since 1960, when most of the coun- 
tries in the region acceded to political independence, African politicians 
have demonstrated the same degree of neglect for these languages. Centers 
for applied linguistics established in the mid-sixties in most of these 
countries (e.g.. Centre de Linguistique Appliquee de Dakar, Institut de 
Linguistique Appliquee at Abidjan) have focused their research on the improve- 
ment of French teaching materials (Turcotte 1981a, P. Dumont 1983), rather 
than on the description of the national languages. Of the six national lan- 
guages recognized in Senegal, only Wolof has benefited of any serious study 
thus far. This situation is paralleled elsewhere in the region. In other 
parts of the continent, in contrast, the study of African languages is at a 
much higher level of development: most of the lingua francas in each coun- 
try have at least a grammar and dictionary of some sort. ■'■■'- This is the case 
with the Ghanaian, Nigerian, Zairean, Kenyan, Tanzanian, Zambian, South Afri- 
can, and Botswanan national languages. 

The paucity, and in many cases the total lack, of language reference 
resources, combined with the lack of qualified language teachers, create a 
vicious cycle in the debate on language policy formulation. It is often 
argued, for example, that African languages cannot be adopted as media of 
instruction, because they (a) are not developed, and (b) require considerable 
investments in the training of teachers. Clearly, as long as African lan- 
guages in Francophone Western Africa remain unstudied and unused as either 
subject or media of instruction, they will never develop the lexicon and 
registers that will permit them to serve adequately the administrative, aca- 
demic, and professional functions they are called to serve. Consequently, 
the language policy in the region will never change. 

To prevent this situation from becoming a self-fulfilling prophesy, the 
cycle must be broken by commissioning over a period of several years the lin- 
guistic study of selected lingua francas in each country so as to lead to 
their eventual teaching as subjects and then adoption as national languages. 
The success of Tanzania, Somalia, and Ethiopia in the use of Swahili, Somali, 
and Amharic, respectively, as national languages has clearly demonstrated that 
the task is not impossible. In order for Francophone African leaders to 
succeed in this endeavor, they must, first, develop the desired political 
will; and second, they must plan carefully their course of action. Currently 
the desired political will is lacking in most of the leaders, although a few 
are moving in this direction. Guinea, Mali, Senegal, and Cameroon are among 
these. Such a change will undoubtedly impact positively on various aspects 
of education: formal and informal, as has been the case in Tanzania, Somalia, 
and Ethiopia (cf. Hinzen and Hundsdorfer 1979, Mochiwa 1984, Adam 1980, Bender 
et al. 1976). 


It cannot be over-emphasized that the study of most of the languages 
identified in Table 8 above was undertaken primary by missionary-linguists 
for religious and educational purposes (cf. Cole 1962, 1971). The demand 
placed on these languages by various educational institutions since the be- 
ginning of this century have necessitated further research that has not only 
increased the number of publications on them, but has also enhanced their 
level of development in a manner which is unparalleled by Francophone West 
African languages. As long as the national languages of this region are not 
used in some major functions or others in education and administration, they 
will never reach the level of lexical and registral development that will en- 
able them to compete against French as media of communication in these areas. 

3.4 Language maintenance and shift . The extent to which a given lan- 
guage or group of languages spreads essentially depends on two major factors: 
the population that uses it, and the function(s) which it is called to serve. 
For example, a language that has a small native-speaker population will not 
spread, i.e., increase the number of speakers, unless it is transmitted to 
the offspring of the said population and/or it is acquired by speakers of 
other languages. Whether the children of the native-speakers of this lan- 
guage and other inhabitants acquire it will depend on the prevailing socio- 
linguistic conditions in the target area. These conditions include the actu- 
al or perceived status of the language; whether it is spoken regularly at 
home and in the community /region; whether it has any socio-economic value 
(e.g., use as the language of market place, employment opportunities, educa- 
tion) . 

A language that is perceived as having no socio-economic value, other 
than cultural, is likely to decline in its currency as its speakers shift to 
one or more other language(s). Ultimately, this decline could lead to lan- 
guage decay and then language death or loss . Such a development may occur as 
a result of built-in natural linguistic and demographic pressures, as indi- 
cated earlier; and could also be the consequence of nationalistic factors as 
expressed in language planning. The subordination of one or more languages 
to another or others as a result of the elevation of a particular language 
to the status of national language has been known as one of the major con- 
tributing factors in language decay and death (cf. Kahane and Kahane 1979, 
Dressier 1982). The decline of Greek and Latin in the Roman Empire in the 
4th and 5th centuries, respectively, Breton in France and Slovenian in Carln- 
thia in modem times, and the death of hundreds of American Indian languages 
in North America in recent history ■'-^ are clear examples of the type of sub- 
ordinating and restrictive language policy that may result from nationalistic 
factors (cf. Brosnahan 1963, Kahane and Kahane 1979, Dressier 1982, Leap 

The decline and eventual decay of a language in cases involving language 
policy formulation, either by a colonizing power or a nationalistic govern- 
ment, come about not only a consequence of social changes, but also as a 
result of a perceived loss in the prestige of such a language. As Dressier 
(1982: 324) aptly observes with regard to language decay: 

(14) Speaker reflect this unfavorable change soclopsychologically by a 
less favorable evaluation of their language. A consequence is a 
socio-linguistically restricted use of their language, which 
results in an impoverished linguistic structure for their language. 


This impoverishment has a feedback on the speakers' sociopsycholog- 
ical evaluation, because the quality for guaranteeing the prestige 
function and the self-identification function (and hence the unify- 
ing/separating functions) of the language has diminished. Aslo the 
sociolinguistically restricted use has a parallel feedback effect. 

Once the speaker of a language have reached this conclusion, viz. that their 
language has little or no socio-economic value and prestige, it is difficult 
to reverse the trend in the decline of the language. The unsuccessful at- 
tempts to raise the status of Irish in Ireland (Macnamara 1970) , Welsh in 
Wales (Lewis 1982), and Yiddish in Central and Eastern Europe (Fishman 1982) 
are illustrative of the difficulties that a "fallen" language encounters. 

African regional or national languages in Francophone Western Africa 
will be headed towards an irreversible decline and ultimately certain death 
unless their current state of neglect is changed. A number of symptons point 
in this direction. First, most of these languages, as indicated previously, 
remain unwritten and unstudied in any serious fashion. Second, they are not 
taught anywhere as subjects of instruction so as to permit youngsters to 
acquire them as second languages in their communities or country. Third, 
many children who are born and grow up in the big cities do not learn their 
mother tongues well, and often not at all; as a result, they become alienated 
from their cultural roots. A consequence of this fact is that the languages 
concerned lose the best segment of the population that would otherwise sus- 
tain them. Fourth and finally, because French is perceived as the prestige 
language and language of socio-economic upward mobility, the teaching and 
learning of any indigenous language at school is seen as a waste of time. 
Accordingly, such languages are devalued from an instrumental perspective. 
This perception can only be changed by an active language policy that allo- 
cates to these national and/or regional languages the kind of functions they 
deserve within the context of a changing African society that must maintain 
its own cultural identity and heritage, while adapting to an increasingly 
interdependent world. 

4.0 Conclusion and Recommendations 

What we have seen in the preceding sections is that the evolution of the 
French colonial language policy in Western Africa was closely related to the 
French colonial objectives: the exclusive use of French as the language of 
education and administration was intended to facilitate the cultural, politic- 
al and economic assimilation of the colonized people so as to ensure French 
domination. A consequence of this language policy, which lasted from 1826 to 
1960, was the neglect of the study of the regional languages in the colonies 
concerned. While there have been other cases of language imposition in the 
modem history of colonialism (cf. Broshnan 1963, Kahane and Kahane 1979, Heath 
1982) , no other colonial power is known to have pursued the ever-encompassing 
assimilation policy practiced by the French in Africa. 

At the beginning of the 19608 the newly liberated countries inherited 
this policy and its consequences in education, both formal and informal. Al- 
most a quarter of century after the advent of political independence in the 
region, the policy has remained largely unchanged; and its negative effects 
are increasingly being felt in education, language development, and the psy- 
chosociological behavior of the children. While a few educators and linguists 


in the region are advocating a change in the language policy, those in power 
want to and have succeeded in maintaining the status quo for various reasons. 

4.1 African languages as tools for development . What is at issue here 
is not only the future of African languages in Francophone Western Africa, 
but also the future of education and its applications in the region. Admit- 
tedly, some African languages will die a natural death as their populations 
decline slowly over the years. In many cases this stage will not be reached 
for several centuries, while in others it may only take a short time. Very 
few linguists and anthropologists encourage the death of languages, but the 
multiplicity of African languages makes any realist welcome such a develop- 
ment for national integration purposes. 

If Africa is to develop both as a geographical and cultural region, it 
must preserve its cultural heritage. And this cultural heritage is in its 
languages, many of which are spoken inter-regionally and internationally. 
Berber, Arabic, Diola, Fula, Bambara/Mandinka, Agni-Baule , Senufo, Ewe, 
Duala, and Wolof , are among such languages in Francophone Western Africa. 
These languages must be preserved and enriched, because they can be the most 
effective tools of personal and national development: they can be used as 
media of instruction, literary development, litercy education, communication 
at the local, regional or national government level, communication for inter- 
nal trade and commerce, etc. Since inhabitants of most the states in the 
region already know one or two of the lingua francas spoken in their country, 
the advantages that they present in serving the above-mentioned functions 
clearly outweigh those offered by French. Once these languages become asso- 
ciated with certain important functions in the society, their prestige will 
rise accordingly and the people will become more motivated to learn them, be- 
cause they represent both societal resources and potential investments. As 
Scotton (1982: 65) correctly observes, 

(15) A resource is something that can be turned to for help or support. 
An investment is something to which people commit themselves for 
future advantage; adding a language is an investment of effort and 


The power or performance of a language as a resource determines 
whether or not a person will seek to learn it. Languages which 
are added are valued not in terms of their specific content,... 
but rather in terms of how they "work" (or do not work); that is, 
they are valued in terms of their interrelation with the other 
societal elements to which learners have access. 

These societal elements include "educational attainment, educational systems, 
political parties, ethnic membership," the individual's linguistic repertoire, 
and the functions to be performed by the acquired language (Scotton 1982: 
63-72). In short, African languages in Francophone Western Africa can become 
valued resources, and therefore tools for personal and national development, 
if other societal conditions are changed. 

It is fallacious to argue that the adoption of selected African lan- 
guages as national languages, and thereby media of instruction, will neces- 
sarily entail the exclusion of French in the communicative functions of the 
nations concerned. This has not happened in Tanzania or Madagascar where 
English and French, respectively, continue to be used as languages of wider 


connnunlcation for international and some national functions. Any objective 
researcher who has been to West and other parts of Africa will recognize that 
French and English are international languages whose role as the languages of 
diplomacy, international trade and commerce, science and technology cannot 
easily be replaced in the near future by African languages, or any other lan- 
guage for that matter. As long as this is true, Africans will continue to 
learn and use them in a complementary fashion with African languages. 
French, therefore, has nothing to fear from a language policy change in 
Francophone Western Africa or elsewhere in the continent (cf. Houis 1971). 
Ethnic rivalry is certainly a major factor to be reckoned with in any lan- 
guage policy formulation in Africa, but it is not an insurmountable difficul- 

4.2 Implications for LP and LP theory . Over a decade ago, Fishman pro- 
posed a typology of language policy decisions in which he suggested that 
factors influencing the type of language policy a developing country may 
adopt can be grouped under three categories: (1) Type A, (2) Type B, and 
(3) Type C (Fishman 1971). 

Type A decisions, according to Fishman (1971: 30), 

(16) ...are those which come about as a result of consensus (at least 
in 'leading circles') that there is neither an over-arching socio- 
cultural past (i.e., no pervasive feeling of unity of history, 
customs, values, or missions traceable into the reasonably distant 
past) nor a usable political past (i.e., no pervasive tradition 

of independence, self-government, hallowed boundaries) that can 
currently serve integrative functions at the nationwide level. 
It is felt by elites in decision-making capacities that there is 
as yet no indigenous Great Tradition (no widely accepted and 
visibly implemented belief-and-behavior system of zndigenously 
validated greatness) that all or most of the inhabitants can im- 
mediately draw upon to make them one people and their country one 

In view of this perception, Fishman suggests, the language policy adopted in- 
variably involves the selection of an LWC (i.e., a language of wider communi- 
cation or international language) . This LWC is often the language of the 
former colonial master. 

In contrast to Type A decisions. Type B decisions characteristically are 
based on a general perception that there exists a Great Tradition in the 
country. This great tradition, according to the author, is based "upon long- 
established socio-cultural unities, and, ...well-established political boun- 
daries as well" (p. 39). In contradistinction to Type A decisions, 

(17) There is widespread consensus — not limited only to elites but 
most consciously and ideologically elaborated by them — that a 
single Great Tradition is available to provide the indigenized and 
symbolically elaborated laws, beliefs, customs, literature, heroes, 
mission, and identity appropriate for nationwide identification 

(Fishman 1971: 39). 

When this type of perception exists, the choice of a language policy leads to 


the selection of an indigenous or indigenized language to serve as the nation- 
al language. Such a policy contributes to and benefits from nationalism at 
its best. 

As for Type C decisions, those are characterized by the existence of 
what Fishman (1971: A5) terms oonfticting or competing multiplicity of Great 
Traditions, none of which is clearly dominant. Fishman observes that, 

(18) Since each of these Great Traditions is numerically, economically, 
and ideologically strong enough to support separate and large- 
scale socio-cultural and political-operational integration, their 
co-occurrence within a single policy makes for rather constant 
internal tension and for nationalistic disunity, particularly in 
the absence of superordinate threat. 

Thus to avoid continued conflict, a multilingual policy is adopted, with a 
foreign LWC serving as the official language at the national level, and the 
indigenous lingua francas of the great traditions serving as regional lan- 
guages (Fishman 1971: 45-48). It is hoped, in this case, that the LWC will 
eventually not only lessen the linguistic rivalries, but would also create a 
sense of national unity and integration. 

Fishman 's typology does undoubtedly describe a number of African 
states. For instance, Tanzania, Somalia, and Ethiopia could be identified, 
to a certain extent, as Type B decision countries. But both Tanzania and 
Ethiopia could also be viewed as Type C nations from a cultural and linguis- 
tic perspective, as there are other major languages and cultures that have 
been subsumed under the present language policies which have favored two of 
the languages in these countries, viz. Swahili in Tanzania, and Amharic in 
Ethiopia. Most of the countries in Africa, and particularly in Francophone 
Western Africa, fall within Type A and Type C decisions. In this regard, 
their adoption of LWC-based language policies is justifiable. 

There are, however, a few countries that are either strictly or predomi- 
nantly monolingual, but have not adopted an indigenous language as a nation- 
al language. Cases in point are Botswana, Swaziland, and Lesotho where 
English continues to serve as the official language. In Francophone Western 
Africa, the focus of this study, Mauritania, Morocco, Algeria, and Tunisia 
have maintained French as the official language, although Arabic is the most 
well-established culture and dominant language. Fishman 's typology cannot 
account for these apparent anomalies. But if it is modified to incorporate 
the colonial legacies factors discussed in this paper and the central role 
that socio-economic factors play in determining the language (s) that individ- 
uals choose to invest in, these anomalies become explanable. That is, the 
countries in which Type B decisions could have been made but have chosen 
instead Type A and C decisions, have done so because of three major consider- 
arations/f actors: (1) colonial legacies (in language policy, education, 
international political links, and economic infrastructure), (2) political 
inertia (resulting from fear of political conflicts and fear of loss of one's 
previleged position of employment), and (3) lack of self-esteem with regards 
to one's native languages. Included somewhere in these factors is the ques- 
tion of misconception concerning the proper place of African languages in 
national and international affairs—a question that can be properly addressed 


only if there is political will on the part of the leadership. This, more 
than anything else, is what has distinguished Ethiopia, Somalia, and Tanzania 
from other African states. 

4.3 Conclusion . We have attempted to show in this study that the cur- 
rent language policies in Francophone Western Africa are the results of both 
colonial and post-colonial policy-decisions. While the intentions of Afri- 
can leaders in maintaining a status quo on the inherited language remain un- 
known, those of the French colonial administrations have been well-document- 
ed. In continuing this policy, African decision-makers have consciously or 
unconsciously espoused the very ideas that debased and colonized them both 
physically and mentally. As long as this mental colonization continues, the 
disastrous effect of the French-based language policy will persist in educa- 
tion and language development for years to come. Over the years many African 
languages will be irretrievably lost without a trace, since they will not 
have been written. The educational systems will continue to be not only 
highly inefficient, but will produce more and more cultural misfits who will 
be comfortable neither in their own culture nor in that of their former 
colonial masters. In consequence, these "intellectuals" will become in- 
creasingly dependent on France and the West for their survival and that of 
their nations. 

To avert this situation. Francophone African states must develop what 
we have termed elsewhere (cf. Bokamba 1981, 1984) comprehensive language 
policies. Such policies involve the objective allocation of functions to 
indigenous LWC and foreign LWC. French, for instance, can continue to serve 
the functions of language of instruction at the university level, language 
of international research, trade, commerce, and diplomacy; while selected 
national languages are given the role of media of instruction in pre-univer- 
sity education, and media of communication at the national, regional, and 
local levels of government. In other words, a multilingual policy is being 
advocated here. To achieve such objectives in the case of Francophone 
Western Africa, the national languages will have to be seriously and careful- 
ly studied. Personnel will have to be trained under government-sponsored 
programs through higher education. With such careful and long-range planning, 
the region will be able to overcome its present difficulties in this area 
with minimal sacrifices. 


I am grateful to Kay Williamson, Braj B. Kachru, and an anonymous re- 
viewer for their comments on an earlier version of this paper. The responsi- 
bility for any errors of facts and/or interpretation are mine alone. 

The focus of the discussion here will be mainly on black Francophone 
Africa, but most of the issues dealt with apply to all former French colonies 
in the continent. 

Specifically, this is what Alexandre (1963: 54) states in this connect- 



Lorsque le Parlement C f ranijais J discuta de I'emploi des langues 
vernaculaires dans 1 'education, ce fut sur la proposition de membres 
metropolitains, et les deputes afviacins s'opposerent d aet emploi, 
en expriment le soupfon qu'il s'agisse d'une manoeuvre destinie a 
empeaher leurs enfants de henefioiev d'une education de niveau identique 
a celle des metropolitains. Plutot qu'un effort en faveur des langues 
vernaculaires, ils exigeaient t 'extension pure et simple, et sur une 
large eohelle, a lever pays, du systeme d'eduaation metropolitain. 
TEmphasis added.) 

This figure includes the two Cameroons: British and French Cameroon. 

Separate figures were not given in the sources consulted for this study. 

4 „ 

An alternative view was proposed by the so-called evolutionist or 

"associationist" school. This school, which included people like Emile 
Durkheira, Joseph Gobineau, Levy Bruhl, Alfred Fouille, and Gustave Le Bon, 
held that Africans and other colonial people were primitive, and hence con- 
siderably different from the French both mentally and culturally. In view 
of this, they maintained, the best approach to use in colonizing and educat- 
ing them was to let them "evolve gradually" towards French civilization by 
providing them rudimentary training in education, political and economic 
organization; rather than by treating them as French equals. This school of 
thought opposed the universal use of French as the medium of education in 
the colonies. For further discussion of this view, see Lokulutu (1982). 

For some discussion of this question, see fcr example Morgenthau (1964), 
Crowder (1968), Calvet (1974), and Lokulutu (1982). 

Guinea opted for immediate political independence in 1958 and rejected 
the offer to join the Franc zone; instead, she established her own monetary 
unit. It must be recognized here, however, that these countries' readiness 
to accept the "French Union" was largely due to economic considerations. 
Many of them were extremely poor and their economic survival depended on 

Other decrees referring to this aspect of the policy included the 
following (Turcotte 1981b: 83-85): 

Article 64, de I'Arret^ No. 1633 du 2 novembre, 1912. 

Aucun livre ni brochure, aucun imprime ni manusarit etrangers cr 

I ' enseignement ne peuvent ^tre introduits a I'ecole sans autorisation 

du Lieutenant-Gouvemeur Csic) . La langue f ran?aise est seule en usage 

dans les ecoles. II est interdit aux maitres de se servir avec leurs 

eleves des idiomes du pays. 

Article 45, de I'Arrete No. 302, du 22 fevrier 1913. 

La langue franijaise dolt etre la seule employee. Les dialectes locaux 

sont rlgoureuseraent interdits. 

In a circular of July 1, 1914, of Governor-General W. Ponty concerning private 
education, it was reiterated that all schools, whether religious, private or 
public, are subjected to the same language policy and objectives In education 


(cf. Turcotte 1981b: 85): 

A cet egard, il est inadmissible que I'enseignement prive, confession- 
nel ou non, echappe a tout contrcle et demeure libre de developper dans 
I'esprit des enfants des tendances qui puissent contrarier nos dcsseins. 

De rrieme que I'ecote officielle, I'eaole priv^e n'a de raison d'etre que 
Qorrme instrument de la cause fvangaise, oomme telle, elle est tenue aux 
memes obligations que I'iaole offiaielle et son existence ne peut ^tre 
toleree que dans la mesure ou elle observe les prinoipes suivants, dont 
vous voudrez bien imposer I'appliaation dans votre colonie. (Emphasis 

Turcotte (1981a) estimates that between 30-A0% of the Ivory Coast's 
population knows French. These estimates appear to be considerably exagger- 
ated, especially in view of the facts (a) that Senegal, the former head- 
quarter of "French West Africa" (i.e., Afrique Occidentale Fran(;aise) and 
the greatest beneficiary of the French education, has a French-speaking 
population of only 11% (Dumont 1983: 324), and (b) that Ivory Coast has a 
much smaller educated population that Senegal. 

9see Eric Lenneberg (1967), The Biological Foundations of Language. 
New York: John Wiley & Sons. 

The University of Illinois' library is one of the top three largest 
university libraries in the United States. Its acquisition in African 
studies is very extensive and representative of the published works on the con- 
tinent. While there may be a few references that the library has not yet 
acquired, the present collection on African languages represents a good 
sample from which certain conclusions can be drawn about the work done thus 
far in the study of these languages. 

It is to be pointed out here that the study of Lingala and Kikongo, 
which are spoken in both Zaire and Congo-Brazzaville, was mainly undertaken 
by missionary-linguists in Zaire (then the Belgian Congo) . Work done on them 
by French and Congolese linguists is a more recent venture, circa 1970s. 


It has been estimated that there were around 500 American Indian langu- 
ages in the United States during the colonial period, but over 250 of these 
are now extinct as a result of U.S. language and educational policies towards 
Indians (cf. Leap 1981). 


ADAM, H.M. 1980. The revolutionary development of the Somali language. 

Occasional paper No. 20. Los Angeles, CA: African Studies Center, 

University of California. 
AFOLAYAN, A. 1976. The six-year primary project in Nigeria. In Ayo 

Bamgbose, ed., Mother tongue education: the West African experience, 

pp. 113-34. London: Hodder and Stoughton. 


AKE, C. 1981. A Political Economy of Africa. Lagos: Longman Group Ltd. 

ALEXANDRE, P. 1963. Les probl^mes linguistiques africaines vus de Paris. 
In J. Spencer, ed.. Language in Africa, pp. 53-59. London: Cambridge 
University Press. 

ANDRZEJEWSKI , B.W. 1980. The implementation of language planning in 

Somalia: a record of achievement. Language Planning Newsletter 6, 1. 

ANSRE, G. 1976. Four rationalizations for maintaining the European lan- 
guages in education in Africa. Paper presented at the International 
Seminar on African languages in education, Kinshasa, Zaire. Mimeo- 
graphed . 

. 1978. The use of indigenous languages in education in Sub-Saharan 

Africa: presuppositions, lessons, and prospects. In James E. Alatis, 
ed.. International Dimensions of Bilingual Education. GURT 1978, pp. 
285-301. Washington, D.C.: Georgetown University Press. 

APRONTI , E. 1974. Sociolinguistics and the question of a national lan- 
guage: the case of Ghana. Studies in African Linguistics, Supplement 
5, pp. 1-20. 

ARMSTRONG, R.G. 1963. Vernacular languages and cultures in modern Africa. 
In J. SPencer, ed . , op. cit., pp. 64-72. 

AWONIYI, T.A. 1976. Mother tongue education in West Africa: a historical 
background. In A. Bamgbose, ed . , op. cit., pp. 27-42. 

BARNES, V.A. 1982. A review of literature on efficiencies in primary edu- 
cation for Sub-Saharan Africa. Paper presented at the A.I.D. seminar 
on education training in Africa, Baltimore, Maryland, March 1982. 

BENDER, M.L, J,D. Bowen, R.L. Cooper, and C.A. Ferguson, eds. 1976. Lan- 
guage in Ethiopia. London: Oxford University Press. 

BOKAMBA, E.G. 1976. Authenticity and the choice of a national language: 
the case of Zaire, Presence Africaine 99/100, pp. 104-43. 

. 1981. Language and national development in Sub-Saharan Africa: 

a progress report. Studies in the Linguistic Sciences 11:1, 1-26. 

. 1984. Language and literacy in West Africa. In R.B. Kaplan, ed. 

Annual Review of Applied Linguistics, vol. 4: Literacy Issues, pp. 
40-74. Rowley, MA: Newbury House Publishers. 

, and J.S. Tlou. 1977. The consequences of the language policies of 

African states vis-a-vis education. In P,A. Kotey and H. Der-Houssikian, 

Eds., Language and Linguistic Problems in Africa, pp. 35-53. Columbia, 

S.C.: Hornbeam Press. 
BOLIBAUGH, J.B. 1972. Educational development in Guinea, Mali, Senegal, 

and Ivory Coast. Washington, D.C.: U.S. Government Printing Office. 
BOTI, M. , M.D. Carelli, and M. Saliba. 1978. Basic education in the Sahel 

countries. Hamburg, W.G. : Unesco Institute of Education. 
BRAY, M. 1981. Universal primary education in Nigeria: a study of Kano 

state. London: Routledge and Kegan Paul. 
CALVET, J.L. 1974. Linguistique et Colonialisme: Petit Trait^ de Glotto- 

phagie. Paris: Payot. 
COLE, D.T. 1961. African linguistic studies, 1943-60. In CM. Doke and 

D.T. Cole, Contributions to the History of Bantu Linguistics, pp. 97- 

107. Johannesburg: Witwatersrand University Press. 
. 1971. The history of African linguistics to 1945. In T.A. Sebeok 

Current Trends in Linguistics, vol. 7: Linguistics in Sub-Saharan 

Africa, pp. 1-29. The Hague: Mouton. 
CROWDER, M. 1967. West Africa under colonial rule, Evanston, IL: North- 
western University Press. 
DENNY, N. 1963. Languages and education in Africa. In J. Spencer, ed., 

op. cit., pp. 40-52. 


DRESSLER, W.U. 1982. Acceleration, retardation, and reversal in language 

decay? In R.L. Cooper, ed . , Language Spread: Studies in Diffusion 

and Social Change, pp. 321-36. Bloomington, IN: Indiana University 

DUMONT, B. 1973. Functional literacy in Mali: training for development. 

Paris: UNESCO, 
DUMONT, P, 1983. Le Francais et las Langues Africaines au Senegal. Paris: 

Editions Karthala. 
FISHMAN, J. A. 1971. National languages and languages of wider connnunication 

in developing nations. In W.H. Whiteley, ed,, Language Use and Social 

Change: Problems of Multilingualism with Special Reference to Eastern 

Africa, pp. 27-56. London: Oxford University Press, 
. 1982. Attracting a following to high culture functions for a 

language of everyday life (the case of the Tshernovits language 

conference in the rise of Yiddish). In R.L Cooper, ed,, op, cit,, 

pp, 291-320, 
HINZEN, H. and V.H. Hundsdorfer, eds . 1979. Education for liberation 

and development: the Tanzanian experience, Hamburg, W.G.: UNESCO 

Institute for Education. 
HOUIS, M. 1971. Anthropologie Linguistique de I'Afrique Noire, Paris: 

Presses Universitaires de France. 
KAHANE, H. and R. Kahane, 1979, Decline and survival of Western prestige 

languages. Language 55:1, 183-98. 
LEAP, W.L, 1981 American Indian languages. In C,A. Ferguson and S.B, 

Heath, eds,, Language in the U,S.A., pp. 116-4A. Cambridge: Cambridge 

University Press. 
LEWIS, E.G. 1982. Movements and agencies of language spread: Wales and 

the Soviet Union compared. In R.L. Cooper, ed., op. cit,, pp. 214-59, 
LOKULUTU, B,M. 1982, Gaullism and Francophone Africa, Unpublished doc- 
toral dissertation, University of Illinois, U-C, 
MACNAMARA, J. 1970. Successes and failure in the movement for the restor- 
ation of Irish. In J, Rubin and B,H, Jernudd, eds.. Can Language Be 

Planned? Sociolinguistic Theory and Practice for Developing Nations, 

pp. 65-94, Honolulu: The University of Hawaii Press. 
MOCHIWA, Z, 1984. Language policy in Tanzania and its impact on education. 

Unpublished paper, Department of Linguistics, University of Illinois, U-C. 
SCOTTEN, CM, 1982, Learning lingua francas and socio-economic integration; 

evidence from Africa. In R.L. Cooper, ed., op, cit., pp. 63-94. 
SPENCER, J., ed. 1963, Language in Africa, London: Cambridge University 


. 1971, Colonial language policies and their legacies. In T.A. Sebeok, 

ed., op. cit., pp. 537-47. 
STUMPF, R. 1979. La Politique Linguistique au Cameroun de 1884-1960: 

Comparison entre les Administrations Coloniales Allemande, Fran9aise et 

Britannique et du Rdle Joue par les Societ^s Missionnaires. Berne, 

W.G. : Peter Lang. 
SYLLA, Y. 1978. Some quandaries of language policy in Senegal: Negritude 

and Francophonie. Paper presented at the 9th Annual Conference on 

African Linguistics, Michigan State University, East Lansing. 
TURCOTTE, D. 1981a. La Politique Linguistique en Afrique Francophone: une 

Etude Comparative de la Cdte d'lvoire et de Madagascar. Quebec: Les 

Presses de I'Universit^ Laval. 
. 1981b. Repertoire Chronologique de la Politique Linguistique en 

Afrique Francophone. Unpublished manuscript. 
UNITED NATIONS. 1983, 1981 Statistical Yearbook, New York: United Nations. 
UNESCO. 1964. Statistical Yearbook 1964. Mayenne: UNESCO, 


UNESCO. 1976. Conference of Ministers of Education of African Member States, 
Lagos, January 27-February 4, 1976: Final Report. 

. 1977. Solving educational problems. Paris: UNESCO/IBE. 

. 1980. Wastage in primary and general secondary education: a statis- 
tical study of trends and patterns in repetition and dropout. Paris: 

. 1983. Statistical Yearbook 1983. Paris: UNESCO. 

WHITELEY, W.H. 1971. Language policies of Independent African states. In 
T.A. Sebeok, ed., op. cit., pp. 548-58. 

studies in the Linguistic Sciences 
Volume 14, Number 2, Fall 1984 

George N. Clements 

Kikuyu, like many other languages of Africa, employs special verb forms 
in the class of syntactic constructions consisting of relative clauses, ex 
situ wh-questions, and focus constructions involving preverbal focused 
elements. This class of constructions shares the further property of being 
subject to a uniform set of constraints governing positions that are 
accessible to extraction rules ("island constraints"). It is argued here that 
the uniquely distinctive characteristic of these constructions is that they 
consist of open clauses, that is, simple clauses whose S-node dominates 
an indexed PRO-form but does not dominate its antecedent. The notion 
binding domain is characterized as a maximal sequence of open clauses. 

1, Introduction, In a number of African languages, widely distributed across the 
continent, we find special verb forms whose occurrence is restricted to a specific set of 
syntactic environments, typically including relative clauses and never including simple 
main clauses This phenomenon is found in languages as linguistically diverse as Hausa. 
Akan. and Swahili. Up to the present time, no linguist has offered a formally coherent 
account of this type of morphological system, although several writers have taken note of 
its existence. Schachter (1973) was the first, to my knowledge, to comment on the 
linguistic interest of a number of highly specific phonological and morphological 
parallels between relative clauses and focus constructions in several unrelated languages 
including Akan. Hausa. and llongo (a Malayo-Polynesian language spoken in the 
Philippines). Other writers have pointed out that wh-questions commonly exhibit formal 
parallels with one or both of these two construction types in certain Bantu languages 
(Myers (1971). Heny (1971). Takizala (1973). Andrews (1975). Epee (1976a.b). Bokamba 
(1976). Elsewhere in Africa, what appear to be similar or related phenomena have been 
identified In a wide variety of languages, including Kpelle (Welmers (1954)). Diola-Fogny 
(Sapir (1965)). Tera (P. Newman (1970)). Fula (Arnott (1970)). Igbo (Robinson (1974)). 
Yoruba (StahIke (1974)). Hausa (R. Newman (1976)). Kanurl (Harrles-Dellsle (1973.1978)). 
and Efik (Cook (1978) and personal communication). This morphological phenomenon is 
not restricted to the African continent, and has been found In such diverse languages as 
Jacaltec (Craig (1977)). Telugu. Malagasy, and Chippewa (Harrles-Dellsle (1973, 1978)), 
as well as llongo, mentioned earlier. 

In the Afrlcanlst tradition, some scholars have used the term "relative tense" to 
describe specially marked verb forms occurring in such environments. However, this 
term Is poorly chosen. Tense is not Involved at all. at least not directly; the distinction 
between "relative" tense forms and simple main-clause forms is not one of tense, but of 
syntactic construction. Moreover, ttie special verb forms in question are most often not 
restricted to relative clauses, but are found in other contexts as well, normally (as 
indicated by the literature cited above) In wh-questions. or in certain types of focus 
construction. The use of the term "relative tense" suggests that the occurrence of these 


forms in relative clauses is primary and that their use elsewhere is secondary, an 
assumption for which little or no evidence is cited in the literature 

This morphological phenomenon is articulated with particular clarity in a Bantu 
language of East Africa. Kikuyu in this language there is a class of syntactic 
constructions (including relative clauses) which are treated as equivalent by several 
independent rules of grammar. In this study we shall attempt to determine exactly what 
this class of constructions has in common Our solution, if correct, will have some 
interest for syntactic theory in general, as it depends upon the recognition of a relation 
of syntactic binding holding between two arbitrarily distant elements, one of which 
may be phonologically null. Such abstract relations are presumably found in most or all 
languages, but only some languages - prominently, those of Africa - encode them in the 
form of special morphological features of the verb. By examining the characteristics of 
syntactic binding relations in Kikuyu. it will be shown that the class of syntactic 
constructions in question is not an arbitrary grab-bag. but crucially involves a notion of 
syntax that I will call the "open clause". 

This class of constructions, which I will refer to mnemonlcally as "class B" due to 
the crucial role played by the syntactic binding relation in accounting for its coherence. 
has the following members: 

(1) a. ex-situ wh-questions in which the questioned constituent 
occurs clause-initially: 

b. preverbal -focus constructions, in which a clause-initial 
constituent is marked as the bearer of focus; 

c. relative clauses 

These constructions behave as a unified set with respect to both syntactic and 
morphological rules. Let us consider the syntactic rules first. Just as many other 
languages. Kikuyu allows sentences to exhibit "long-distance" or unbounded dependencies 
between an overt noun phrase and a position in the sentence from which an identical noun 
phrase is understood to have been extracted (i.e.. moved or deleted), in English, for 
example, we find sentences such as Who do you think I said Bill saw? in which the 
object of saw has been extracted, and a dependency exists between that position and the 
initial question word who In English, such extractions are subject to certain constraints 
originally termed "island constraints" by Ross (1967), which account for the ungrammati- 
cality of sentences like Who do you think Mary knows a person who saw? 
(illustrating a violation of Ross's Complex NP Constraint) or Who do you think Mary 
wondered who ja»'.^ (illustrating a violation of the Wh-lsland Constraint (Chomsky 

A similar set of constraints upon long-distance dependencies exists in Kikuyu. In 
Kikuyu, however (which contrasts with English in this respect), such constraints may be 
circumvented in a fairly free manner by the use of "resumptive pronouns" in the position 
of a nonsubject gap. We shall examine relevant constructions just below. Our first 
point, then, will be that the constructions of class B form a unified set with respect to 
the island constraints of Kikuyu. 

If we turn now to Kikuyu morphology, we find that a certain number of rules of word 
formation, affecting the morphological shape of verbs, apply in this same class of 
constructions. Certain of these rules apply individually in other construction types as 
well, but they apply jointly only in the constructions of class B. The rules in question 


are the following: ^ 

(2) a. a verbal suffix consisting of the downstep element /7 is deleted. 
When not deleted, this suffix is realized to the right of the first 
complement following the verb (such as the direct object) if 
there is one, and otherwise at the end of the verb In most 
tenses, tonal forms of the verb are different in other respects 
as well. 

b. If the extracted element is a class 1 (typically, singular human) 
subject, then the usual subject prefix (SP) of this class, /a-/ 
(with underlying high tone, indicated by the acute accent), is 
replaced by the class 1 pronominal prefix (PP) lo-l (with under- 
lying low tone, unmarked in transcriptions). This latter prefix 
is used elsewhere in the concord system for pronouns and for 
noun qualifiers other than adjectives 

c. In negative clauses, the negative prefix l-X'\-l is replaced by 

These points can be briefly illustrated as follows. (3a) is a simple declarative 
sentence, and (3b-d) are the corresponding class B constructions as defined in (1) above: 
(3b) is an ex-situ wh-question. (3c) is a focus construction, and (3d) is a relative 
clause. ^ 

(3) a. KariokT a-'tem-ire mo-te' 'Kariuki cut a tree' 

SP-cut-T CP-tree 

b. no.o o-tem-ire mo-te? 'Who cut a tree?" 
FP-who PP-cut-T 

c. ne Karioki 6-tem-'ire mo-te 'it's Kariuki (that) cut a tree" 
FP PP-cut-T 

d. mondo o-ria 6-tem-'ire mo-te 'the person (that) cut a tree' 
person PP-DEM PP-cut-T 

The following abbreviations will be used in grammatical glosses: SP=subject prefix, 
PP"pronominal prefix, T-tonse/aspect prefix or suffix, CP"nominal class prefix, 
DEM=demonstrative modifier, FP=focus particle. The latter is an element occurring 
preposed to nouns and verbs which indicates the scope of focus, and whose basic shape is 
/ne/: it combines with the wh-pronoun o 'who' to form its clause-initial alternant no.o. 
as we see in (3b). 

(3a) illustrates the simple main clause form of the verb, formed in this tense (the 
simple past completive, which Is used in all affirmative past tense examples except when 
otherwise indicated) by prefixing the appropriate SP and suffixing the tense/aspect suffix 
-ire. As it is not a class B construction, this sentence does not display the special 
properties indicated in (2a, b). In particular, as required by (2a). the verbal downstep 
suffix r I occurs in its expected position to the right of the following complement.'' and 
the verb shows the "normal" tone form appropriate to a simple main clause. (The 
downstep internal to the verb originates in the preceding noun, and has nothing to do 
with the downstep suffix just mentioned.) Furthermore, as is required by (2b). we find 
the SP /a-/. 


(3b) illustrates the subject question corresponding to (3a) Both of the properties 
described in (2a, b) are illustrated here First, we find a new tonal form of the verb, 
which does not have the clause-final downstep suffix. Second, the SP /a-/ has been 
replaced by the PP lo-l . 

(3c) is the subject-focus construction corresponding to (3a), and illustrates the same 
properties as (3b) (The fact that the low-toned PP lo-l bears surface high tone in this 
example is explained by a phonological rule of Downstep Displacement, which converts a 
tone sequence of the form H'l" (where L"= one or more L tones) to one of the form HH"'; 
this rule has affected the first two syllables of the verb, which bear low tones at an 
earlier stage of derivation, and explains the downstep occurring internally in the verb. 
As the tone rules of Kikuyu are fairly complicated, I will not attempt to give an account 
of surface tones In the remainder of this study, except where directly relevant to the 
discussion. Interested readers are referred to Clements and Ford (1977) for a fuller 
account of phrase- 1 eve I tone rules.) 

Finally, (3d) is the subject relative corresponding to (3a), and shows the same 
morphological properties as (3b. c). 

The replacement of the negative prefix /-ti'-/ by /-ta-/ can be illustrated by the 
examples given in (4). (4a) is an affirmative sentence, (4b) the corresponding negative, 
and (4c) the subject question corresponding to (4b). ^ 

(4) a. ka-ana -ja-'tem-ire mo-te' 'The child cut a tree' 

CP-child SP-cut-T CP-tree 

b. ka-ana -ja'ti'na-tem-'a mo-te 'The child didn't cut a tree' 


c. no.o o-ta-na-tem-a mo-te? 'Who didn't cut a tree?" 
FP-who PP-NEG-T-cut-T 

In (4a) and (4b) we find the class 12 SP -ya-. in concord with the cl. 12 (diminutive) 
subject noun kaana. As we see in (4b), the main-clause form of the negative prefix is 
-//-, while -na - is the tense prefix required in this negative tense (see Armstrong 
(1967), Barlow (1964) for discussion of verb morphology). In (4c), which has no.o 'who' 
as the subject, we find that the class 1 PP -o - appears as the verbal prefix, indicating 
that we have a "class B" environment. In confirmation of this, the negative prefix shows 
the alternant -ta- (with different consequences for tonal realization). 

We have given an overview of a number of syntactic and morphological phenomena 
which demonstrate the unified nature of class B constructions. In order to see exactly 
what the members of this class have in common, we will now examine their structure 
more closely. At the same time we will give examples confirming our as yet unsupported 
claim that these constructions are subject to a set of island constraints similar in nature 
to those found in English and other languages. 

Before doing so. however, it would be appropriate to remark on certain phonological 
rules affecting vowel sequences which, though irrelevant to the syntactic analysis as 
such, affect the surface form of Kikuyu words, often making their morphological structure 
difficult to decipher. The processes relevant to the examples in this paper are listed 


(5) a. au 

b. ao 

c. ae 

d. 00 

These processes are illustrated below. {6a) presents a table of the subject prefixes (SP) 
for all six persons, illustrated with a paradigm of the verb -tern - 'cut' in the simple 
past completive tense As this stem is consonant-initial, the prefix vowels show up in 
their basic form. (6b) presents corresponding paradigms of three vowel-initial verb 
stems for comparison, illustrating the effect of the rules of (5), as well as several others 
which will not be involved in our later examples. Tones are not marked. 

(6) a. sg. pi. -tem- 'cut' 

totem ire 






1. p. N- 



2. p. 0- 



3. p. a- 





-on- 'see' 

















-uif- say 
njuTfire tuutire 
uuYirc muuifirc 
oiYire moifire 

In addition to rules affecting vowel quality, other rules affect vowel quantity. For 
example, a long vowel is shortened when occurring in the same syllable as another vowel 
Keoo -» eo, etc.). All these rules apply both within words and across word boundaries. 
In order to make the examples more intelligible, I have not indicated the effect of such 
rules in the subsequent transcriptions in this paper. The reader may convert the tran- 
scriptions into actual spoken forms by applying the rules of (5) and the vowel shortening 
rule, with appropriate tonal adjustments. Further discussion of vowel coalescence 
processes can be found in Armstrong (1967, chapter 3). 

2. Constraints on Extraction Rules, fx-situ wh-questions are formed, as we have 
seen, by proposing a question word such as no.o 'who' and leaving a gap in the position 
from which the questioned NP has been extracted. That the dependency between the 
question word and the rejated gap is unbounded is suggested by sentence (7a) below, and 
many similar sentences: 


(7) a. no.o o-jw-eciiri-a Ngo-je a-uf-ire ate Kama.u a-on-'ire? 
FP-who SP-T-think-T SP-say-T that SP-see-T 

"Who do you think NgugT said that Kamau saw?' 

Here the dependency extends two clauses down, relating no.o 'who' and the object of 
a on 'ire 'saw'. 

When the extracted element occurs in an island, as illustrated In (7b-d), the result 
is ungrammatical: 

(7) b. "no.o Kama.u a-6n-'ire mo-ndo o-rea 6-ring-'ire? 
FP-who SP-see-T CP-person PP-DEM PP-hit 

"Who did Kamau see tne person (that) hit?' 


c. *n6,o NgoY6 a-eciri-ne hiihi no.o o-on-iTe? 

SP-wonder-T PP-see-T 

'Who did NgugT wonder who saw?' 

d. "no Kama u a-on-'i're Kaanake mbere ya'' 

SP-see-T front of 

'Who did Kamau see Kanake in front of' 

(7b) illustrates a violation of Ross's Complex HP Constraint, in which a questioned 
element has been extracted from within a relative clause: the ill-formed dependency here 
holds between an element occurring outside the relative clause (the question word no.o ) 
and a gap occurring within it (the extracted object of oring' iri 'hit') (6c) illustrates 
a Wh-island Constraint violation, hers, the illicit dependency holds between the question 
word no o and an object gap occurring in a clause that contains another extraction, 
involving the subject (6d) illustrates the fact that phrases consisting of a form of the 
nominal modifier /-a/ of followed by a NP constitute islands; a pronominal NP in such a 
construction cannot enter into a dependency with an element occurring outside, explain- 
ing the ungrammaticality of (7d). 

Interestingly, the ungrammatical sentences in (7b-d) can be substantially improved by 
placing a "resumptive pronoun" in the position of the extracted element. Thus we find 
that (7b', c) are semi-acceptable for some speakers consulted, and (7d') is acceptable for 
all speakers (meanings are identical to the corresponding examples in (7b-d)): 

(7) b'. no.o Kama.u a-6n-'ire mo-ndo o-rea 6-m6-ring-'ire? 
FP-who SP-see-T CP-person PP-DEM PP-OP-hit-T 

c'. no.o Ngoife a-eciri-rie hiihi no.o o-mo-on-i're? 
SP-wonder-T PP-OP-see-T 

d'. no.o Kama.u a-on-'i're Kaanake mbere ya-ke? 
SP-see-T front of-POSS 

What I am terming "resumptive pronouns " in these examples consist of the 3rd. sg. OP 
(object prefix) /-mo-/ in (7b', c'), and the 3rd. sg. POSS (possessive modifier) /-ke/ in 
(7d'). These formatives are identical in shape to the corresponding "free" pronominal 
formatives, in all respects: that is, there is no formal distinction in Kikuyu between 
"resumptive" pronouns and other pronominal forms bearing the same set of morpho- 
syntactic features. 

In addition to the ex-situ wh-questions discussed up to this point, in-situ 
wh-questions occur in Kikuyu as well. In which the question word occurs in the "logical" 
position of the questioned element: 

(8) Kama.u a-6n-'ire o? 
SP-see-T who 
'Who did Kamau see?' 

Such questions do not exhibit the morphological effects listed In (3). In such questions 
the focus particle ne cannot be prefixed to the question word. A further feature of 
In-situ wh-question formation is that only non-subjects may be questioned in this way; 
subjects are obligatorily extracted, as we see by applying the morphological criteria In 
(2) (see example (3b) above, for which no corresponding in-situ question exists). 
According to the speakers whom I have consulted, in-situ and ex-situ wh-questlons 
are nearly or entirely synonymous. 


We find an entirely parallel set of constraints on extraction when we turn to 
preverbal focus constructions. The following examples illustrate focus constructions 
involving the subject, the object, and a locative complement respectively; ° 

(9) a. ne Kama.u o-6n-'ire Kaanake 

FP PP-see-T 

'It's Kamau (that) saw Kanake' 

b. ne Kaanake Kama 'u a-on-'i're 

'It's Kanake (that) Kamau saw' 

c. ne mbere ya mo-te o6-rea Kama.u a- on- ire Kaanake 

front of CP-tree PP-DEM SP-see-T 

'It's in front of that tree that Kamau saw Kanake' 

In all such cases, just as with ex-situ wh-questions, the focus particle ne is 
obligatorily prefixed to the initial NP. In terms of discourse function, the initial NP 
constitutes the "focus" of the assertion in contrast to the rest of the clause, which 
presents the background information. 

Focus constructions such as those illustrated in (9) must be distinguished from topic- 
al izations, which are distinct both in formal and semantic/pragmatic terms. In topical iz- 
ations, the proposed element is not preceded by the focus particle ne. and the following 
portion of the sentence is not characterized by the morphological processes listed in (2). 

(10) mbere ya mo-te o6-rea Kama.'u a-6n-'ire Kaanake' 
front of CP-tree PP-DEM SP-see-T 

'In front of that tree Kamau saw Kanake' 

In contrast with (9c). the postverbal downstep suffix is not deleted in (10), and appears in 
its normal position after the object, where it preserves the final high tones of Kaanake 
from deletion (cf. note 4). 

Focus constructions exhibit the same set of island constraints that hold in ex-situ 
wh-questions, as illustrated below: 

(11) a. ne Kama.u n-gw-e'ciiri-a Ngoye a-UY-ire ate o-on-ire Kaanake 
FP SP-T-think-T SP-say-T that PP-see-T 

'It's Kamau (that) I think NgGgT said (that) saw Kanake' 

b. *ne Kaanake Kama.'u a-on-'ire mondo o-rea 6-ring- ire 

SP-see-T CP-person PP-DEM PP-hlt-T 
'It's Kanake (that) Kamau saw the person (that) hit" 

c. "ne Kaanake Ngo^ye a-eciri-rie hiihl no.o o-on-i're 

SP-wonder-T FP-who PP-see-T 

'It's Kanake (that) NgugT wondered who saw' 

d. " ne Karioki Kama.'u a-6n-'i're Kaanake mbere ya 

SP-see-T front of 

'It's Kariuki (that) Kamau saw Kanake in front of 

Furthermore, just as with wh-questions, the island constraint violations can be circum- 
vented by the insertion of resumptive pronouns in the position of the extracted element. 
In the case of focus constructions, however, the result is fully acceptable for all 


dUb' ne Kaanakc Kama.'u a-6n-'ire mondo o-rea 6-m6-ring-'ire 

SP-see-T CP-person PP-DEM PP-OP-hlt-T 

c'. ne Kaanake Ngo've a-eciri-rie hiihi no.o o-mo-an-i're 

SP-wonder-T FP-who PP-OP-see-T 

d' ne Karioki Kama.'u a-6n-'i're Kaanake mbere ya-ke 

SP-see-T front of-POSS 

A further parallel between focus constructions and wh-questions is that the formal 
counterpart of in-situ wh-questions exists for focus constructions: that is, constructions 
in which the focused element remains in its basic position In such cases, the morpho- 
logical processes listed in (2) do not take place: instead, the placement of focus on an 
element following the verb is indicated by other morphological features of the verb.^ 

Let us turn finally to relative clauses. Relative clauses in Kikuyu are normally 
constructed by placing a demonstrative modifier (DEM) such as /-reaV after the head 
noun, inflected with the appropriate PP. The relative clause follows this modifier 
immediately with no intervening relative pronoun or complementizer, and with no 
modification of the verb other than by the processes listed in (2). '^ An example 



(12) oyo ne [ |op 'mo-ndo o-rea [o o-6n-'ire Kaanake ] ] 
this FP CP-person PP-DEM PP-see-T 

This is the person (that) saw Kanake' 

In respect to island constraints, relative clauses exhibit the same range of properties as 
the constructions discussed earlier. This is Illustrated below: 

(13) a. mo-ndo o-rea n-gw-e'ciiri-a Ngoife a-uTf-ire ate Kama.u a-on-'ir'e 

CP-person PP-DEM SP-T-think-T SP-say-T that SP-see-T 

'the person (that) I think NgugT said (that) Kamau saw' 

b. "mo-ndo o-rea Kama,'u a-6n-'ire mo-ndo o-rea 6-ring-'ire 

SP-see-T PP-hit-T 

'the person (that) Kamau saw the person (that) hit' 

c. "mo-ndo o-rea Ngoife a-eciri-rie hiihi no.o o-on-ire 

SP-wonder-T FP-who PP-see-T 

'the person NgugT wondered who saw' 

d. "mo-ndo o-rea Kama,'u a-on-'JTe Kaanake mbere ya 

SP-see-T front of 

'the person (that) Kamau saw Kanake in front of 

As in the case of focus constructions, the use of resumptive pronouns renders the last 
three examples fully acceptable to all speakers consulted: 

(13) b'. mo-ndo o-rea Kama.'u a-6n-'ire mo-ndo o-rea 6-m6-ring-'ire 
SP-see-T PP-OP-hit-T 

c'. mo-ndo o-rea Ngo-je a-eciri-rie hiihi no.o 6-mo-on-ire 

SP-wonder-T FP-who PP-OP-see-T 
d", mo-ndo o-rea Kama.'u a-6n-'iTe Kaanake mbere ya-ke 

SP-see-T front of-POSS 


Let me summarize the discussion so far. We have examined a certain range of 
syntactic constructions, which we have called "class B", and found that they share a 
significant number of properties: 

- they provide a context for a set of morphological rules (2) affecting the 
form of the main verb: 

- they Involve unbounded dependencies between an overt NP and a follow- 
ing extraction site: 

- when the extraction site contains a phonological ly null element the un- 
bounded dependencies are subject to a uniform set of constraints on 

- violations of these constraints can be minimized or avoided by the use 
of resumptive pronouns in place of nonsubject gaps. 

Most traditional theories of grammar provide no explanation for this constellation of 
properties. Such theories require us to list, for each rule or constraint in question, the 
set of syntactic constructions providing their context. No explanation is given as to why 
exactly the same set of constructions should behave as a single set with regard to a 
number of independent rules of grammar. Our problem, then, is to determine the identity 
of the property, call it "Property X". that is shared by all these constructions. 

3. Analysis. As the preceding discussion has shown, "gaps" and resumptive 
pronouns stand in a relation of complementary distribution within the set of acceptable 
sentences exhibiting unbounded dependencies in Kikuyu. Gaps occur grammatically in 
"accessible" positions, that is positions that lie outside of Islands, and only there. 
Resumptive pronouns occur grammatically in "inaccessible" positions, that is, within 
islands, and only there. The fact that resumptive pronouns do not occur grammatically in 
"accessible" positions is easily confirmed by taking note of the unacceptability of 
sentences like the following: 

(14) "ne Kaanake Kama.'u a-mo-6n-'i're 
FP SP-OP-see-T 

'It's Kanaka (that) Kamau saw' 

(compare the corresponding grammatical sentence In (9b)). This relation between "gaps" 
and resumptive pronouns distinguishes the syntax of Kikuyu from that of languages like 
English, in which island constraint violations cannot normally (in standard varieties) be 
circumvented by the use of resumptive pronouns. 

In order to account for this range of facts, a small number of assumptions will be 
made regarding basic phrase-structure properties of Kikuyu syntax. These assumptions 
will be justified in the following discussion, but it should b>e added that they receive 
further motivation from aspects of Kikuyu syntax that cannot be dealt with here. I 
propose the following phrase structure rules, as a first approximation, to account for the 
constituent structure of full clauses (S"), the complementizer position (COMP). and the 
focus position (FOC); 

(15) a. S' -► COMP S 

b. COMP -» Comp (FOC) 

c. FOC -• ne (NP) 

The NP generated as an optional constituent by (15c) will be further subclassified by the 
feature [±WH]. If the feature [+WH] is assigned to this NP. (15c) will generate 


ex-sltu wh-questlons: if the feature [-WH] is assigned, it will generate focus con- 
structions If NP is not expanded at all. it will generate grammatical sentences such as 
the following, in which the FP ne has been moved to preverbal position by a clitici- 
zation rule: 

(16) Kama u ne-a-on-'i're Kaanake 

'Kamau saw Kanake' 

Notice further that if FOC is not expanded in (15b), we will not generate either ex-situ 
wh-questions or focus constructions, since the initial NP of these two construction types 
must occur under the immediate domination of the FOC node. An analysis of this sort 
provides a maximally straightforward account of the range of construction types under 
consideration here, and will be adopted for the purposes of this study. 

The present analysis accounts for a range of further observations. For example, the 
rules of (15) generate only one FOC position per S-clause , and thus predict that no 
S'-clause may have both a preposed wh-question word and a proposed focus constituent. 
This is correct, as the ungrammaticality of examples such as the following shows: 

(17) "no.o ne Kama.u 6-ring-'ire? 

FP-who FP PP-hit-T 

■Who is it Kamau (that) hit?' 

As long as one or the other of these elements is not preposed, the sentence is grammati- 
cal, as (18) shows: 

(18) ne Kama.u 6-ring-'ire o? 
FP PP-hit-T who 
■It's Kamau (that) hit who?' 

Here the question word o does not occur in the clause-initial FOC position, and so the 
sentence Is grammatical. Notice that since (18) is synonymous with the Intended reading 
of (17), the ungrammaticality of (17) cannot be accounted for on semantic grounds. Nor 
can the ungrammaticality of (17) be due to island constraints of some sort, since the 
corresponding sentence with a resumptive pronoun in place of the gap is totally 

(19) "no.o ne Kama.u 6-m6-ring-Mre? 


A further correct prediction of the present analysis is that two focused NPs may not 
occur at the head of the same clause. This prediction follows, like the previous one, 
from the fact that the rules of (15) generate only one clause-initial FOC position. 
Accordingly, sentences like the following are ungrammatical: 

(20) "ne Kaanake ne Kama.u 6-ring-'ire 
FP FP PP-hit-T 

'It's Kanake (that) its Kamau (that) hit' 

As soon as one of these NPs is placed in a non-FOC position, the sentence becomes 


grammatical (cf. the examples in (9a. b)), but then, of course, the second NP is no longer 
marked for focus. We can show once again that the ungrammaticality of (20) does not 
involve an island constraint violation by taking note of the ungrammaticality of the 
corresponding sentence with a resumptive pronoun: 

(21) "ne Kaanake ne Kama.u 6-m6-ring-'ire 

FP FP PP-OP-hit-T 

For the same reason, two wh-questlon words may not occur in preposed position in 
the same clause: 

(22) "no.o no.o o-on-ire? 

FP-who FP-who PP-see-T 

'Who saw who?' (lit. 'who did who see?') 

Once again the resumptive pronoun is impossible. Furthermore the ungrammaticality of 
(22) cannot be due to semantic considerations, since some speakers, though rejecting 
(22). find (23) acceptable in the same meaning: 

(23) no.o o-on-ire o? 

Finally, it will be observed that this analysis provides only one source for the FP 
ne in any clause, namely as an element introduced under the immediate domination of 
FOC within COMP. It therefore follows that no clause may contain more than one 
occurrence of the particle ne . This prediction, too, is correct: prenominal ne and 
preverbal ne mutually exclude each other, and so all sentences containing both of them 
are ungrammatical: 

(24) "no.o ne-a-on-i're Kaanake? 

FP-who FP-SP-see-T 
Who saw Kanake?' 

Just as in the preceding case, the ungrammaticality of sentences like (24) cannot be 
explained on semantic grounds, that is, as involving some sort of incompatiblity between 
the semantics of prenominal and preverbal ne. To show this, it is necessary to mention a 
further type of ex-situ wh-question in Kikuyu, In the case of multiple embeddings in 
which the extracted wh-element binds a gap occurring at least two levels of embedding 
down, the wh-question word need not occur at the head of the highest clause, i,e, initial- 
ly in the sentence, but may occur in a synonymous reading at the head of any intermedi- 
ate clause. Thus the following three sentences, which differ syntactically only in the 
location of the question word no.o. are grammatical and synonymous: 

(25) a. no.o 6-Yw-eciiri-a Ngoife a-uK-ire ate o-on-ire Kaanake? 

FP-who SP-T-think-T SP-say-T that PP-see-T 

b. 6-Yw-'eciiri-a no.o Ngoye a-UY-ire ate o-on-ire Kaanake? 


c, 6-Yw-'eciiri-a Ngoye a-uy-ire ate no.o o-on-ire Kaanake? 

'Who do you think NgugT said saw Kanake?' 

However, the freedom of occurrence of such "wandering" question words is subject to the 
restriction that no question word may appear at the head of a clause containing preverbal 

ne (26b) is ungrammatical for this reason: '^ 

(26) a. ne-ko Ngoife a-uif-ire ate Kama.u ne-a-on-'i're Kaanake'? 
FP-where SP-say-T that FP-SP-see-T 

b. "Ngoye a-ut-ire ate ne-ko Kama.u ne-a-on-'lTc Kaanake'? 


c. NgoYe a-UY-ire ate Kama.u ne-a-on-'JTe Kaanake 'ko? 

all meaning: 'Where did NgugT say (that) Kamau saw Kanaka?' 

Since both the ex-situ wh-question (26a) and the in-situ wh-question (26c) are 
grammatical. (26b). which has the same intended meaning as both of these, cannot be 
ill-formed on semantic or logical grounds. Its ungrammatical ity can, however, be 
explained by the fact that the set of rules (15) generates only one occurrence of the FP 
ne per clause, and so cannot generate (26b). 

Relative clauses will be generated by the following base rule: '^ 

(27) NP -► NP S 

This rule characterizes relative clauses as right complements of NPs. and assigns them 
the categorial status of simple clauses, S, as opposed to complementizer-introduced 
clauses, S'. The choice of S as opposed to S' is motivated by the fact that neither 
complementizers, relative pronouns nor (preposed) focused NPs can occur initially in a 
relative clause in Kikuyu. Were rule (27) to expand the relative clause as S' rather than 
S, thereby allowing a FOC position to be generated in clause-initial position, these 
exclusions would be unexplained, and would call for a set of special statements to 
account for them. The following sentence demonstrates the inability of focused NPs to 
occur at the beginning of a relative clause: 

(28) "mo-ndo o-rea ne Kama.u 6-ring-'ire 
CP-person PP-DEM FP PP-hit-T 

'the person (that) it's Kamau (that) hit' 

The explanation for the ungrammatical ity of (28) cannot be that the dependency between 
the head of the relative clause and the gap involves an island constraint violation, since 
the corresponding example with a resumptive pronoun is also ungrammatical: 

(29) "mo-ndo o-rea ne Kama.u 6-m6-ring-'ire 


Nor can the ungrammatical ity of (28) be due to any semantic restriction on the occur- 
rence of focused NPs within relative clauses, since focused NPs occur freely in relative 
clauses as long as they do not occur in the highest (matrix) clause. This variety of facts 
is explained most simply by the syntactic solution proposed In (27). 

On the basis of these assumptions, we may return to the property of "class B" con- 
structions remarked on earlier, namely that in acceptable sentences in Kikuyu. "gaps" and 
resumptive pronouns stand in a relation of complementary distribution. This situation is 
summarized below: 


(30) a. gap In accessible position: grammatical 

ne Kaanake Kama.'u a-6n-'ire 

'it's Kanaka (that) Kamau saw' 

b. gap in island: ungrammatlcal 

"ne Karioki Kama.'u a-on-'ire Kaanake mbere ya 

'It's Kariuki (that) Kamau saw Kanake in front of 

c. RP in accessible position: ungrammatlcal 

"ne Kaanake Kama.'u o-mo-on-'iVe 
(same meaning as (a)) 

d. RP in island: grammatical 

ne Karioki Kama.'u a-6n-'ire Kaanake mbere ya-ke 

(same meaning as (b)) 

To account for this distribution we will postulate the following rule of Colndexing arxJ 

(31) Colndexing and Deletion 

NP [g X [+PRO] Y ] 

SO: 1 2 3 4 

SC: (a) CO index 1 and 3 

(b) delete the terminal node of 3 if 3 occurs In an 
"accessible" position with respect to t 

Part (a) of the SC accounts for all unbounded dependencies of the type under consider- 
ation in this study. It might well prove that such colndexing is better carried out in some 
other way, e.g. by a separate rule, or by the free generation of indices in the base and 
the provision of filtering principles to rule out ill-formed linkages. As the study of 
coindexed structure as such is not the subject of this study, we will leave this issue 
open, and turn to the part (b) of the rule, which is of more interest here. This part of 
the rule carries out the deletion of "accessible" pronominal elements which are bound, by 
coindices, to a full UP to their left. It must be assumed that coindices of the sort under 
discussion here (those involved in the characterization of unbounded dependencies 
between two elements in "class B" structure) are formally distinguishable from coindices 
that might be proposed for other purposes, such as the statement of relations of free 
anaphora. As most current syntactic theories draw such a distinction in one way or 
another, I will suppose that this assumption is unproblematical. 

(31) provides a full account of the facts summarized in (30): index-bearing 
pronominals will fail to be deleted (and will survive in the form of resumptive pronouns) 
if ar>d only if they occur in islands with respect to their antecedents 

4. Binding Domains in Kikuyu. We have now provided a basis for understanding 
the fact that the "class B" constructions of (1) behave in a uniform way with respect to 
island constraints: they do so by virtue of the fact that they are all derived by a single 
rule. (31). It is not yet clear, however, why thp morphological rules listed in (2) should 
be able to apply to just this class of constructions. The problem is that under a reason- 
ably constrained view of syntactic theory, rules of grammar do not have access to the 
derivational history of their ir>puts. except to the extent that previous rules have left 
indirect effects of their operation in the form of traces, indices, and the like. 


Given the analysis developed so far. however, it should be clear that the con- 
structions of class B share a common structural property as a result of the operation of 
(31), namely the formal relation of coindexing that holds between two NPs under the 
conditions stated m the SD of (31). 

We may characterize this structural property in the following way. Let us define the 
notion open clause as any clause S whose root node dominates an indexed pronominal 
[+PRO] and does not dominate its (coindexed) antecedent. Schematically, given a 
structure of the following form containing an unbounded dependency between a pronom- 
inal element [+PRO]j and its antecedent NPj, we may say by virtue of the definition that 
S^ through S^ are open clauses, while other Ss in the structure (i.e. those dominating S^ 
or dominated by S^) are not In particular, S^+j is not an open clause; 



Intuitively, an open clause is one containing an unbound variable of a particular type: an 
indexed pronominal. 

Under the analysis developed so far. all the members of "class B" syntactic con- 
structions have the following form: 

(33) NPj [s ... [+PRO]| ... ] 

They are thus all instances of the schema (32), abstracting from the presence of the 
lower clause 5^+^. The context for the application of the morphological rules of (2) may 
now be stated as "main verb of an open clause." 

But what of more complex structures involving multiple embedding? The definition of 
open clause given above makes two predictions regarding such structures: first, that the 
main verbs of clauses dominating NP: will not exhibit the morphological characteristics 


in (2), and second that the main verbs of clauses not dominating the pronominal element 
[+PRO] will also fail to exhibit these characteristics. Both of these predictions are true, 
as we shall now see. 

We have already seen examples that allow us to test the first prediction, in the 
earlier discussion of "wandering" wh-question words The sentences of (25) contain 
three levels of embedding, as follows; 

(34) [S, ... [S2 .. [S3 ...]]] 

In all cases, the pronominal "gap" (subject of the lowest verb) is in S 3. These sentences 
differ among themselves in the location of the antecedent. The structure of these ex- 
amples is schematized in (35): 

(35) (25a): NPj [5^ [g^ [53 I*PRO], ... ] ] ] 

(25b): (s, NPi [s^ [33 ■ [*PRO]| ... ] ] ] 

(25c): [g^ [s2 NPi [3^ ... [+PRO]i ...]]] 

By the definition given above, the open clauses are Sj, So and S3 in (25a), So and S3 
in (25b), and only S3 in (25c) Accordingly, we expect that if any of the morphological 
rules of (2) are applicable, they will affect only ttie open clauses, and thus fall to apply 
to S| of (25b) and S, and S2 of (25c). 

The morphological evidence confirms our expectation. It will be recalled that in 
some tenses (such as the current past completive, illustrated here), special tonal forms 
are used for verbs in "class B" constructions. If we now examine the tonal structure of 
the main verbs in Sj and S2 in (25), we find the "special" tonal form ofweciiria in 
(25a) and the "normal" tonal form <?/•»»' Wc//>/a in (25b. c). Similarly, we find the 
"special" form aiijire (with H tone influence on the following complementizer ate ) in 
(25a, b) and the "normal" form auyirc (with no H tone influence on ate ) in (25c). The 
lowest verb, oonire, has its "special" form throughout. This is entirely in conformity 
with our analysis Main verbs have their "special" forms if and only if they occur in open 
clauses. The tonal variation found in sentences such as (25) is a direct consequence of 
the fact that rules of tonal morphology have access to the notion "open clause", as 

characterized above 


Let us consider now the second of the two predictions, concerning the morpho- 
logical behavior of clauses satisfying the description of 3^.,+ , in schema (32). We expect 
that such clauses will not show the characteristics of (2), even when the conditions under 
which these rules apply are otherwise satisfied. Examples such as (36) show that our 
expectations are again fulfilled; 

(36) a No.o Kama u a-er-'ire Kaanake atn o-tem-ire mo-te? 
FP-who SP-tell-T that PP-cut-T CP-tree 

'Who did Kamau tell Kanake (that) cut a tree?' 


b, No.o Kama u a-er-'ire ate Karioki a-'tem-ire mo-te'? 

Who did Kamau tell (that) Kariuki cut a tree?' 
c No 6-Yw-eciiri-a o-UY-ire ate Kamau a-'tem-ire mo-te'? 
SP-T-thlnk-T PP-say-T SP-cut-T 

'Who do you think said (that) Kamau cut a tree?' 

In (36a). the wh-word no.o binds a gap (in our analysis, a coindexed. phonologically 
null pronominal) in the subject position of the lowest clause. Hence, no clause satisfies 
the description of S_+| in (32); all clauses are open clauses, and we expect them to 
exhibit open-sentence morphology. In (36b), on the other hand, no.o binds a gap in the 
object position of the main clause, and in (36c) it binds a gap in the subject position of an 
intermediate clause. Thus the lowest clauses of both (36b, c) satisfy the description of 
S^+l in (32), and we do not expect them to exhibit open-sentence morphology. 

Our analysis is confirmed in two ways. Notice first that in (36a) the lowest verb, 
'cut', exhibits the effect of rule (2b), which replaces the SP /a-/ with the PP Iq-I . 
(36b, c), however, retain /a-/ More importantly, we see that the postverbal downstep 
has been deleted by rule (2a) in the lowest clause in (36a), while it is retained in 
(36b. c), preserving the final high tones of /770/d' in both instances. In other words, 
downstep deletion has taken place in (36a). where the lowest clause Is an open clause. 
but not in (36b. c), where it is not. 

The evidence discussed in the last three paragraphs provides particularly strong 
evidence for the central role of the notion "open clause" in Kikuyu grammar, and so for 
its importance as a category of linguistic theory. Without this notion - or some notion 
similar to It - we would be unable to explain the full range of data discussed in this 
paper in a convincing way. We would have to resort, in one way or another, to a listing 
of the environments in which the morphological rules of (2) apply. 

The evidence from Kikuyu demonstrates that certain rules of grammar may refer to the 
notion "open clause" in their structural description. The fact that in Kikuyu, as in many 
other languages, the property of being an open clause is registered on the verb rather 
than on some other category suggests that certain mechanisms may "encode" this property 
on the verb in the form of an abstract feature, which is realized in surface structure (or 
not) according to the morphology of each particular language. 

We may finally characterize the notion binding domain as a maximal chain of open 
clauses S] S ^ meeting the structural conditions stated in (32). within which morpho- 
logical or phonological rules may apply in some languages under conditions specific to 
these languages, as in the case of Kikuyu. The operation of these rules provides an overt 
morphological encoding of the domain of discontinuous dependencies within a given sen- 
tence, crucially depending on abstract aspects of hierarchical clause structure, such as 
the presence of phonetically null elements. 



This is a slightly expanded version of a paper presented at the 10th Annual Confer- 
ence on African Linguistics at the University of Illinois (Champaign-Urbana). April 5-7. 
1979, under the title "An Unbounded Deletion Analysis of Wh-questions in Kikuyu." As the 
title suggests, my original presentation gave greater attention to the arguments in favor of 
an analysis deleting the coindexed [+PRO] element, rather than moving it by cyclic wh- 
movement. As the issue of movement vs. deletion seems less topical today than it did in 
1979, I have deemphasized this aspect of the discussion here, but the arguments in favor 
of a deletion analysis remain strong, and are implicit throughout the present version. See 
Bresnan and Grimshaw (1978) for a framework similar to the one assumed here. In other 
respects, I have made no attempt to revise the analysis originally presented in 1979, but I 
have added a few references to more recent work, where relevant, in the footnotes. 
Research for this study was carried out in Nairobi. Kenya in October-December 1978. and 
was supported by grants from the Harvard Graduate Society and the Marion and Jasper 
Whiting Foundation, to whom I would like to express my appreciation. I would also like to 
thank Professor Mohamad Abdulaziz of the University of Nairobi for his generous assistance 
in the course of my research in Kenya. In a later stage of preparation of this paper I bene- 
fited from helpful discussions with Frank Heny and Annie Zaenen. I would finally like to 
thank two anonymous SLS reviewers for several suggestions leading to improvements in the 

The variety of Kikuyu reported on in this paper represents the speech of university 
Students enrolled at the University of Nairobi. I would especially like to thank NgugT wa 
Karenge (Murang'a District). John QTtau (Murang'a District), and Wambui Kaire (Nairobi 
District, formerly of Kiambu District) for their help Ali linguistic examples cited in the 
text were checked with each of these individuals, who were generally very consistent in 
their judgements (disagreements are noted in the text). Most examples were later re- 
checked with Mrs. Lillian MwanTki. of Nairobi, during her stay in Cambridge. Massachu- 
setts in 1979-80. 

^ A fuller account of the phonology and morphology of Kikuyu nouns and verbs can be 
found In Clements (1984). The "relative" tense forms of that paper are the "special" forms 
briefly described in (2a. b). The distribution of the postverbal downstep is somewhat more 
complicated than is indicated in (2a). in ways that do not bear directly on the analysis of 
this paper. For example, as first observed by Ford (1976), the downstep suffix is absent In 
all negative tenses, where there is no reason to suppose that It forms part of the urwJerlying 
tonal morphology of verbs at all. It is also absent in yes/no questions, as the result of a 
general rule deleting all downsteps In such questions. On the other hand, the downstep 
suffix appears in all ne- tenses (tenses formed with the focus particle ne ), regardless of 
whether the verb In question occurs in a "class B" construction or not; for an example, see 
(26) below. It should be added, finally, that in the fuller analysis of Clements (1984). the 
downstep is not actually deleted in "class B" environments, but simply fails to be inserted 
by the rules of inflection. 

^ Transcriptions In this paper follow the system used In Clements (1984) Tonal 
diacritics are as follows: /a/ high tone, /a/ falling tone, /a/ rising tone, /a/ low tone 
(unmarked). The raised exclamation point /7 represents downstep. Wh-question 
intonation, affecting the last two syllables of a sentence only, is represented here by the 
question mark /?/. Vowel sequences within a word are single syllables except where 
separated by a dot. e.g. /au/ is one syllable and /a.u/ two; the tone of each syllable is 


marked on the final vowel. Transcriptions do not reflect the phonological effect of syllable 
fusion across word boundaries. In glosses, hyphens isolate inflectional affixes: deriva- 
tional morphology is not indicated except in the case of the noun class prefix. Finally, 
standard Kikuyu orthography is used in the transcription of proper names, where /", u. e , and 
o have the phonetic values [e. o. e, o] respectively, and vowel length is not indicated. 

'' Although the downstep element is unpronounced, its presence at the end of sentences 
like (3a) may be verified not only by extrapolation from the rules that account for its 
presence sentence- internally, but also from its effect upon immediately preceding high 
tones and rising tones. Such tones are normally lowered to low in sentence-final position, 
but retain their underlying value on the surface if a sentence-final downstep follows. In 
the case of (3a), for example, the rising tone on the final syllable of mote indicates that a 
downstep follows, since if no downstep were there, it would be realized as a low tone 
Conversely, since the lexically-determined rising tone of mote surfaces as a low tone in 
(3b-d). we know that no downstep follows 

^ As observed in note 2, the failure of the sentence-final downstep to appear in (4b, c) 
is due to the morphological fact that no negative tenses have the downstep suffix, no 
matter what syntactic context they appear in 

° The word-final high tones in aon'ire in this sentence, as well as in the final 
verbs of (7b', c'), (llb',c'). and (13b', c') below, might appear to be unexplained exceptions 
to the rule mentioned in note 4 that lowers word-final high tones when not followed by 
downstep. There Is a phonological explanation for this apparent exceptionality, however. 
As shown in Clements (1984), verbs in tenses formed without tense prefixes, including the 
current past completive illustrated in these examples, have a floating low tone suffix which 
predictably blocks the operation of two tone rules, one of which is the tone lowering rule 
in question. 

Most speakers whom I consulted In the preparation of this study were unable, after 
consideration of relevant examples, to find any semantic or pragmatic distinction between 
corresponding in-situ and ex-situ questions, although one speaker felt that only ex- 
situ questions were appropriate when the person to whom the question Is addressed is 
asked to identify one or more individuals or objects from a previously-established set. 

^ The vowel of the PP is lengthened with the demonstrative /-reaV to express the 
sense 'that (yonder)' (cf. Benson (1964;395), Armstrong (1967:12), Barlow (1960:36)). When 
this form Is used as a definitizer, as in the relative clause examples elsewhere In this 
paper, the PP vowel has its normal length. 

^ Namely, absence of the preclitic FP. and certain tonal modifications (Clements 

''^ This analysis is in agreement with that of Barlow (1960). who writes (p. 56): 
"Kikuyu relative sentences are expressed without a relative pronoun. (...) When the 
antecedent is definite, one of the demonstrative adjectives uyO. urTa, ucio. &c.. must 
be associated with It. (...) In general propositions the demonstrative is optional." 

The occurrence of the FP as an apparent "copula" in this sentence is unrelated to 
the fact that a relative clause follows. 


'^ Final high tones and rising tones are realized as falling tones under question 



Under certain assumptions, for instance those of Jacl<encloff (1977), a phrase- 
structure rule such as (27) is ill-formed due to the fact that an identical category appears 
on both sides of the rewrite arrow Nothing of consequence hangs on this feature of our 
analysis, however. Further investigation might show reason to assign the NP to the right to 
a lower rank, within a theory of X'-syntax 

No attempt is made here to provide a formal account of the conditions under which 
a coindexed pronominal is "accessible", i.e. deletable under the operation of (31b) Such 
an account would be of considerable interest, but would take us far beyond the more modest 
goal of the present discussion, which is to explain the array of data in (30) in the context 
of the problem of characterizing "class B' constructions. 

^ Here I will point out an unexplained anomaly which arises under the tonal 
analysis assumed elsewhere in this study. In (25b. c). one would expect the postverbal 
downstep suffix to appear after verbs displaying their "normal" tonal forms However, the 
downstep suffix has apparently been deleted In all three cases, as well as in the highest 
verb in (26c) One possible hypothesis, which I have not been able to investigate, is that 
the downstep suffix is deleted sentence-internally after any verb that falls within the 
logical scope of a wh-question word. What the exact nature of this rule is, and whether it 
holds for all speakers, must await further study. 

'° In recent work. Zaenen (1983) has formalized a notion of syntactic binding 
related to the one presented here within the framework of Lexical-functional Grammar. 
Showing that it accounts for a wide range of linguistic phenomena, both morphological and 
syntactic, Bergvall (1983) examines further aspects of syntactic binding phenomena in 
Kikuyu, showing that an apparent class of island constraint violations involving inanimate 
antecedents is explained by an independent rule of inanimate object-prefix deletion; she 
makes the interesting observation that in-situ wh-questions are subject to the same set of 
island constraints as their ex-situ counterparts. 


ANDREWS. A. 1975. Studies in the Syntax of Relative and Comparative Clauses. 

Unpublished Ph.D. dissertation. M.I.T 
ARMSTRONG. L, E. 1967. The Phonetic and Tonal Structure of Kikuyu. London: 

published for the International African Institute by Dawsons of Pall Mall. 
ARNOTT, D. W. 1970. The Nominal and Verbal System Of Fula. London: Oxford 

University Press. 
BARLOW, A R 1960. Studies in Kikuyu Grammar and Idiom. Edinburgh: Blackwood and 

BERGVALL, V. L. 1983. "WH-questions and Island Constraints in Kikuyu: a Reanalysls." 

In J. Kaye, H. Koopman, D Sportiche, and A Dugas. eds.. Current Approaches to 

African Linguistics, vol.2 Dordrecht: Foris Publications. 
BOKAMBA, E. G 1976 Question Formation in Certain Bantu Languages Unpublished 

Ph.D. dissertation. University of Illinois. 
BRESNAN, J. and J. GRIMSHAW. 1978. "The Syntax of Free Relatives in English." 

Linguistic Inquiry 9. 331-391. 
CHOMSKY. N 1977. "On Wh-Movement " In P. W. Culicover, T Wasow and A Akmajian. 

eds.. Formal Syntax. New York; Academic Press. 


CLEMENTS. Q N, 1984 "Principles of Tone Assignment in Kikuyu." In Q. N. Clements 

and J Goldsmith, eds , Autosegmental Studies in Bantu Tone. Dordrecht: Foris 

CLEMENTS, G N. and K C Ford 1977. "On the Phonological Status of Downstep in 

Kikuyu." In Q. N. Clements, ed.: Harvard Studies in Phonology, vol 1 Reprinted in 

D. L. Goyvaerts, ed , Phonology in the 1980's Ghent: Story-Scientia, 1981 
COOK. T. L. 1978 "The Formation and Use of the Positive Relative Verb Forms in Efik 

and the Structure of the Relative Clause " Xeroxed manuscript. University of Leiden. 
CRAIG. C G 1977 The Structure of Jacaltec Austin: The University of Texas Press. 
EPEE. R 1976a Generative Syntactic Studies in Duala Unpublished Ph. D dissertation, 

Cornell University. 
EPEE, R. 1976b "On Some Rules That Are Not Successive Cyclic in Duala." Linguistic 

Inquiry 7.1, 193-8. 
FORD, K. C 1976. Tone and Intonation in Kikuyu. Unpublished manuscript. University of 

HARRIES-DELISLE. H. 1973/1978. "Contrastive Emphasis and Cleft Sentences." Stanford 

University Working Papers on Language Universals 12, 85-144. A revised version 

appears in J. H. Greenberg. ed.. Universals of Human Language, vol. 4: Syntax. 

419-86. Stanford: Stanford University Press. 
HENY. F. 1971. "Focus and Interrogative in Bali " Unpublished paper, presented at the 

2nd Annual Conference on African Linguistics. Los Angeles. 
JACKENDOFF. R. 1977. X' Syntax: a Study of Phrase Structure. Linguistic Inquiry 

Monograph No. 2. Cambridge, Ma.: MIT Press. 
LUKAS, J. 1937. A Study of the Kanuri Language. London: Oxford University Press. 
MYERS, A. 1971. "On the Similarities between Interrogatives and Emphatics in Kikuyu." 

Proceedings of the 2nd Annual Conference on African Linguistics (Studies in African 

Linguistics, Supplement 2). 11-17. 
NEWMAN. P. 1970. A Grammar of Tera. Berkeley: University of California Press. 
NEWMAN. R. M. 1976. "The Two Relative Continuous Markers in Hausa." Studies In 

African Linguistics. Supplement 6. 177-190. 
ROBINSON. J. O. S. 1974. "Focus-presupposition and WH-questions in Igbo." In E. 

Voeltz, ed.. Proceedings of the 3rd Annual Conference on African Linguistics. 

Bloomington: Indiana University Press. 
ROSS, J. R. 1967. Constraints on Variables in Syntax. Unpublished Ph.D. dissertation, 

SAPIR, J. D. 1965. A Grammar of Diola-Fogny. West African Language Monograph no. 3. 

London: Cambridge University Press. 
SCHACHTER. P 1973. "Focus and Rel at ivizat ion." Language 49.1. 19-46. 
STAHLKE. H. 1974. "Pronouns and Islands in Yoruba." Studies in African Linguistics 5.2, 

TAKIZALA. A. 1973. "Focus and Relativization: the Case of Kihung 'an." In J. P. Kimball, 

ed.. Syntax and Semantics, vol.2, 123-148. New York: Academic Press. 
WELMERS, W. E. 1964. "The Syntax of Emphasis In Kpelle." Journal of West African 

Languages 1.1, 13-26. 
ZAENEN, A. 1983. On Syntactic Binding. Linguistic Inquiry 14.3, 469-504. 

studies in the Linguistic Sciences 
Volume 14. Number 2, Fall 1984 


Chet A. Creider 


Data from videotaped conversations in a number of East African languages 
and in English are used to establish the existence of a preference in the African 
interactional systems for the use of verbal feedbaci< by listeners in contrast 
to a preference for the use of non-verbal (gaze, head nod) feedback in English. 
This difference is related to different understandings of the meaning of gaze 
behavior and to differences in preferred mutual postural orientation for 
conversation, deixis, and speaker turn length between the English and the African 

Work by sociologists on the organization of conversation has produced evidence for 
the existence of a variety of structures in terms of which conversational interactions 
function. The turn-taking system is discussed in Sacks, Schegloff and Jefferson's 1974 
Language article, "A simplest systematics for the organization of turn-taking in 
conversation." A later article in Language by the same group of authors discusses the 
repair system (Schegloff, Jefferson and Sacks 1977). The system organizing the beginning 
and ending of conversations was discussed in lectures by Sacks under the label, "the overall 
structural organization of conversation." Due in part to the recency of this work, there 
has been little attempt to study the organization of conversation from this perspective 
in languages and cultures other than English. One exception, Moerman 1977, presents 
evidence that Tai conversation, in at least one aspect, the repair system, functions almost 
exactly like English. 

Because conversation is ultimately at the basis of all social interaction, and because, 
in fact, a surprisingly strong case can be made for the genesis in conversation of "abstract" 
knowledge and more generally for the interactional nature of the creation of knowledge^, 
it is of considerable importance to know if the conversational systems of all human cultures 
operate in ways which are basically the same. 

Most anthropologists' instinctive answer to this question is 'impossible,' but a reasonable 
argument can be made for suspecting that the above surmise is indeed correct. On an 
empirical level the list of similarities between Tai and English conversations provided 
by Moerman is very impressive. On a less specific level, the kinds of systems found to 
operate in conversation operate with reference to sets of structural positions in 
conversations (transition-space for turn-taking, 'repair-initiation opportunity space' for 
repairs (Schegloff, Jefferson and Sacks 1977, 375)) which may be presumed to exist (i.e. 
to be definable and wtilizable by interactors) in all conversations, regardless of the language 
being used. The systems in which these structural positions are utilized, moreover, appear 


to have as a primary objective (or 'conspiracy') ensuring that a speaker has a 'fair chance' 
to speak in the face of a variety of potentially disturbing factors (such as competition 
between two listeners for next turn). A 'fair chance' for a speaker can be regarded as 
a kind of minimal requirement for meaningful conversation (anywhere), and hence the 
possibility appears of language-independent conversational systems. In general then, it 
can be argued that to the extent that the subtle ways in which conversation is organized 
are due to the nature of interaction itself (i.e. to the functional requirements which any 
system which is to organize conversation must be responsive to) then culture- and 
language-specific conventions may not be found. 

In this paper 1 make no attempt to address in an overall way this question of 
language/cultural differences in interactional system. 1 present some empirical evidence 
for the existence of language-specific differences in the preferred mode (communicational 
channel) of realizing particular interactional functions (or effecting particular kinds of 
interactional behaviour) in conversation. 1 then argue that these differences have impact 
on other interactional subsystems and more generally on the kind of interactional work 
that is accomplished with conversation. 

More specifically, the difference in preference for eye contact in interaction between 
English and a number of East African languages is studied. The massive presence of the 
difference is first established, and then the existence of an alternative mechanism is 
described. This mechanism itself allows certain differences in the sequential organization 
of conversation, and thus will be seen to shed light on the major issue outlined above. 

Data from videotaped conversations from a number of East African languages (Kipsigis, 
Luo, Gusii, Samburu) establish the existence of a preference in the African interactional 
systems for the use of verbal feedback by listeners in contrast to a preference for the 
use of non-verbal (gaze, head nod) feedback in English. The existence of this latter 
preference for English has been noted in a number of studies, e.g. Kendon 1970, 
Birdwhistell 1971. Godwin 1981 is especially relevant as he notes that eye contact is 
a necessary precondition for the initiation of an utterance directed at another. My own 
work, reported below, fully supports this finding, which by now must be regarded as fully 

The preference for verbal feedback by listeners in the African conversational systems 
is related to differences in other aspects of the interactional systems: different 
understandings of the meaning of gaze behavior (supportive and interested for English, 
hostile and rude for the African systems), preferred mutual postural orientation for seated 
conversation (facing or angled for English, side by side for African), deixis (speaker and 
listener 'points of view' are the same in the African systems, but mirror images of each 
other in English), speaker turn length (shorter in African conversations), use of and response 
to negative questions (answer more strongly presupposed in English conversations), etc. 

Although they are not examined here, I also claim that there exist other 
language/cultural differences in the use of particular communicational channels to effect 
particular communicational functions and that these differences have an impact throughout 
the rest of the set of interactional systems which organize conversation for particular 
cultures. Some examples are the use of syntactic resources versus the use of 
stress/intonation for signalling marked focus, and the use of hand movements versus the 
use of syntactic resources as cohesive devices (in the sense of Halliday and Hasan 1976). 

1 do not of course want to argue that each language is a thing-unto-itself. One of 
the major outcomes of work into a variation-based theory of language universals (see 
for example E.L. Keenan 1976), has been the realization that a framework in which attention 
is paid to specific differences is a viable way, paradoxically, of investigating that elusive 


creature. Language, with a capital L, or in our case, Conversation, with a capital C. I 
suggest as worthwhile tasks for the next stage of investigation the accumulation of further 
data such as those presented in this paper, and then the elaboration of a general theory 
of the interactions of the behavioral components themselves in implementing the various 
functional prerequisites of conversational interaction. 

Some of my own data now follow. For four conversations, one between two Canadian 
English speakers (male and female, university students, VT 46), one between two Kipsigis 
elders (VT 1-1), one between two Kipsigis secondary school students (one male and one 
female, VT 27-3), and one between two Luo elders (VT 20-1), I recorded the presence or 
absence of gaze directed at the other interactor every thirty seconds (this period is 
considerably longer than the average turn length and so eliminates any intra-utterance 
effect that might be present). The results are as follows. 



VT 46 (English) 
VT 1-1 (Kipsigis) 
VT 27-3 (Kipsigis) 
VT 20-1 (Luo) 













in percentages: 


VT 46 
VT 1-1 
VT 27-3 
VT 20-1 













The magnitude of contrast involved here would be duplicated with any set of 
conversations from my data. Note especially the very high (50%) incidence of mutual 
gaze in the English conversation and the extremely low incidence rates for all of the African 
conversations. Note also the close agreements between interactors within a conversation, 
VT 20-1 being the only exception in this respect. 

Another way of assessing the utilization of non-verbal versus verbal feedback is to 
compare the frequency of use of verbal listener feedback in the conversations. In the 
English conversation the number of 'yehs' and similar utterances which functioned to 
indicate comprehension but which were not elicited by a preceding interrogative utterance 
was 28 out of a total of 266. In the Kipsigis conversation (VT 27-3) the number of such 
utterances in 266 turns was 73 (p<0.01, binomial test). 

Verbal and non-verbal strategies have different implications for some fairly obvious 
matters such as turn length. Given the opportunity to ensure that the listener is attending 
as a speaker speaks, the speaker need not pause as frequently 'in the course of a turn', 
but may, provided the listener is attending, construct lengthy turns. Not utilizing non-verbal 
feedback it is to be expected that the African conversants would create more frequent 
opportunities for listener verbal feedback, and hence that average turn length would be 
shorter in these conversations. 


In the videotape 46 English conversation the average number of words per turn was 
13.2 while in the comparable Kipsigis conversation (27-3) this average was 4.8 
(p<0.01, t_-test). These figures were based on a 73 turn sample in the case of the English 
and a 108 turn sample in the case of the Kipsigis. The extremes of the distributions are 
revealing: there were 37 (34.2%) one word turns in the Kipsigis conversation and 9 (12.3%) 
one word turns in the English conversation. At the other end of the distributions, there 
were 3 Kipsigis turns of more than 20 words: 21, 22, and 28. There were 13 such English 
turns including 46, 48, 50, 78, 81 and 85 word turns. 4 

I want to emphasize that the differences in eye contact are differences in preference 
only. That is, both systems include both mechanisms for feedback. The use of non-verbal 
behaviour increases dramatically in African conversations which are not dyadic. This 
may be understood as follows: In all systems an increase in the number of speakers means 
an increase in competition for the floor. It is clear that non-verbal means are more 
efficient than verbal ones for this purpose (they take up less sequential space, do not involve 
potential loss of the floor through giving it temporarily to the listener, etc.). This means 
that with dyadic conversations we have two possibilities. Either there is more competition 
for the floor in the English dyadic conversations and less in the African dyadic ones, or 
there are cultural conventions which operate to keep a system adapted to competition 
operative in circumstances under which it is strictly speaking not necessary. That is, 
English dyadic speakers could use less non-verbal feedback but don't. Regardless of which 
possibility is actually the case it is clear that cultural conventions do operate in this 
micro-arena even when they are strictly speaking unnecessary. Even if the second 
possibility is not the case, and the first one is, then either English speakers in dyadic 
contexts maintain, by convention, a level of competition which is in a functional sense 
unnecessary, or African speakers, by cultural agreement, operate with lowered levels 
of competition in dyadic contexts. ^ We may further note that while non-verbal behaviour 
is more functional under conditions of greater floor competition, it is in general a less 
preferred alternative: it involves an extra modality, restricts operating freedom, etc. 

It will be clear from the foregoing that African speakers and listeners are oriented 
in conversation to expect frequent and regular verbal feedback from listeners, and that 
furthermore this means that regular gaps are provided in the production of utterances 
by speakers in the expectation that listeners will fill them in, and moreover fill them 
in with such productions as will not take the floor away from current speaker. This may 
also be looked at from the point of view of the listener: whose utterances are accorded 
a floor of their own and are secure from being overlapped. An interesting linguistic 
difference may be related to this, one which is regularly noted, but has never been 
understood in any way: listener responses in English are subject to very severe mandatory 
ellipsis constrictions, while the same types of utterance in the African conversations do 
not even allow ellipsis (VT 46): 

G: Aah, Brescia [local college], where're you from? 

P: Tronuh [Toronto] 

Note that utterances such as 

P: I'm from Toronto 

are regularly non-occurrent in this position. Contrast the Kipsigis (VT 5-2b): 

B: ee, iyapuu ano oo? 'Where do you come from?' 

A: apunuu komosito isinee lekemaani. '1 come from down this way.' 

Examples such as this are of regular occurrence in the African conversations. 

A somewhat more subtle linguistic difference which may result from the 'expanded 
floor' of the African systems is the following (VT 46, English): 

G: You mean, like, you didn't come to do Anthropology? 

P: No. Like, I wanted Anthropology, I definitely wanted it, but ... (continuation not 



and later in the same conversation 

G: Where? Not here? 

P: No. I'd like to do the U.B.C. 

In each of the above cases P's nos would be realized as 'yes'es' in the African languages, 
as shown below (VT 27-3, Kipsigis): 
A: mm, ki. oo, maatoopwaan ak loorit-is? 

B: ee, matakeepwaan ak loorit, keepwaan een kaasiit aap aeeng. 

A: And didn't you come in the lorry? 
B: yes, we didn't come in the lorry, we came on Tuesday. 

In the English questions the negative in the response is extracted from and repeats 
a portion of the question, but in the African questions the response indicates that the 
question is taken to be of the general form: "is (not(S)) true?" The English question 
presupposes the truth value of the questioned negative assertion, but the African sequence 
does not. and rather, includes it in the domain of the question (c.f. Pope 1973). 

This example is in fact typical of many of the differences between the English and 
the African conversational materials. The English ones usually presuppose more (more 
precisely, the speaker implicates the listener in more presuppositions than do the African 

Given the preferred mutual postural orientations mentioned previously, we may note 
that the 'natural' spatial conversational arena for the two participants in a conversation 
is in front of them and not between them for the English speakers. This has some very 
interesting consequences, 1 believe, for deixis systems. In the African systems it follows 
that the speaker's and listener's orientations are identical, but that in the English system, 
speaker's and listener's points of view are not identical. Now consider how English speakers 
describe the following situation: if there is an object, say a ball, between ourselves and 
another object, say a table, we say that the ball is in front of the table. If the ball is 
beyond the table, we say that it is behind the table. That is, in each case, we take the 
point of view of the table in such a way that v;e assume it is 'facing' us (note that, as 
discussed in a delightful paper by John Kimball (1968), tables and the like don't have faces 
which are given in nature). In contrast, in each of the African languages discussed, the 
ball would be described in the first instances as being behind the table and in the second 
as being in front of the table. This is exactly the opposite of English usage! Speaking 
metaphorically it is clear that African speakers do not assume a point of view for the 
table which is opposite their own. Rather they assume that the table has a point of view 
which is the same as theirs. This difference is exactly what we would expect on the basis 
of the conversational behaviour. 

Another point related to deixis: the linguist A.L. Becker has discussed the use of the 
first person in English as a kind of unmarked point of reference, and contrasted this with 
the use of the second person in politeness systems in a number of Southeast Asian languages. 
He uses the term 'reverse deixis' for this phenomenon. For example, a letter might begin, 
"How are things here?," where the "here" means, in our terms, "there". That is, the 
addressee's point of view is taken (Becker 1979).^ 

With reference to body motion, Birdwhistell (1971) has observed that inwardly directed 
pointing movements are associated with both the first person personal pronouns and with 
the proximal demonstratives (this, these), while outwardly directed pointing movements 
are associated with both third person pronouns and the distal demonstratives (that, those). 


We may add to Birdwhistell's pronoun list the second person pronouns which also are 
associated with outwardly directed movements and note that the contrast is then first 
versus non-first person in association with proximal versus distal deixis (an additional 
argument in support of Becker's reverse deixis concept for English). 

A striking difference between the African and the English conversations is that although 
the African languages have full repertoires of personal pronouns (including both bound 
and free forms), and although pointing movements are common in association with spatial 
deictics (which also are well-developed), pointing movements are not associated with 
pronominal reference. 

Furthermore there is a difference in the way the pointing movements are used in the 
African conversations. Specific locations are pointed to, but there is not a regular and 
systematic use of a contrast between here/me and there/other. That is, the system is 
as often gradient as equipollent (in Trubetskoy's terminology). 

Also, needless to say, there is no justification for the association of the 'meaning' of 
the proximal demonstrative with 'near to the speaker' and the medial demonstrative with 
"near to the listener" with these African languages. The terms have a primarily spatial 
reference in the first instance, and that spatial reference is to areas relatively near/away 
from both the speaker and the listener. 

A point related to the preceding: it has been noted that American English speakers 
customarily converse at distances which are relatively far apart (with reference to Arabic 
or Latin American norms (Watson 1970). While I suspect that part of this has to do with 
the fairly major importance of interactional leg use in this culture (Sarles 1977), I think 
that part is also attributable to the use of the space between interactors in our culture. 
In African cultures this space need not be used in interaction, and so it doesn't matter 
if it is filled up (with the interactors' bodies). In fact it appears that in the three African 
cultures where 1 studied the effect of distance in detail experimentally (reported in 
Creider 1977), distance makes no difference whatever. There is a tendency to prefer 
a closed position when convenient, but a striking feature of the African scene is the huge 
distances over which people routinely carry on extended conversations. 

A somewhat more abstract matter related to the previous discussions: there is an 
interesting contrast in the style of presentation of discursive arguments (and possibly 
of the conception of these arguments — I do not believe that the conception of an argument 
is an event separate from its interactional realization). Spoken English conversations 
in my experience tend to utilize either-this-or-that type contrasts with great frequency 
in argumentation. The African conversations, however, routinely assume as an unmarked 
case that the number of arguments or facets or whatever of an issue exceeds two. This 
is typically given body motional realization by hooking the fingers of one hand over the 
individual fingers of the other hand as points are made.^ 

We have moved speculatively fairly far afield from our original starting point. Let 
me close with the following set of observations which brings us back to the beginning. 
We started out by noting that the African conversants did not make systematic use of 
interactional gaze behaviour, and in particular did not use it to provide listener feedback. 
This was in marked contrast to the usage patterns of English speakers. We may go further 
and note that there is a general tendency in the African interactional systems to orient 
to the spoken part of the conversation exclusively. This also is in marked contrast to 
the English systems, where a person's nonverbal behavior can be and often is oriented 
to, and where in fact a variety of disciplines have sprung up whose existence depends 
on the reality of this orientation. Regardless of its validity, Birdwhistell's off-quoted 


statement that in "normal two person conversation, the verbal components carry less 
than 35% of the social meaning; more than 65% is carried on the nonverbal band,"^ is 
one that can only be made in a culture where nonverbal behaviour is in fact a significant 
part of interpersonal interaction. Similarly the work on verbal deception, 'leakage', and 
in general the possibility of saying one thing with one's words and another with one's body 
which is done by psychologists studying nonverbal behaviour (e.g. Ekman and Friesen 1974), 
only makes sense in a culture such as ours which is oriented to nonverbal behavior. 

In contrast the African cultures are rather singularly unoriented to the nonspoken 
word as an indicator of interactional meaning. Note that in the African cultures, interactors 
do find themselves in situations where for reasons of politeness and respect they must 
say one thing and perhaps mean another, and they are quite aware that this is something 
which they will encounter in the speech of others. But when it comes to detecting 'true' 
intentions, it is speech which is oriented to and not nonverbal behaviour.^ 


^ An earlier version of this paper appeared in 1980 as ERIC document ED 200 053. 
I am grateful to Dan Jorgensen for pointing me to the sources cited in note '^ and also 
for discussing many of the points in this paper with me. He is not responsible for any 
errors of representation or interpretation I make with respect to the New Guinea cultures. 

2 This is, I believe, one of the major significances of Sack's 1972 article on puns, and 
is also in Goodwin 1981. 

^ A low incidence of gaze use has been reported (without data) by Scheflen (1972) 
for Black North Americans, but with no understanding that this was a piece in a functioning 
interactional system which could be realized in other modes. 

^ A check on the number of morphemes per word for the Kipsigis conversation shows 
a figure of 1.77. This corresponds closely enough to Greenberg's (1960) 1.68 value for 
English that I haven't made any correction for this factor. 

^ Note that some interesting questions are located here. Dyadic conversations, with 
respect to the question of competition for the floor, are qualitatively different from triadic, 
etc. conversations. From the point of view of the listener, in a dyadic conversation the 
only 'competition' is current speaker, who occupies (it will be recalled from Sacks, Schegloff 
and Jefferson 1974) a privileged position in the turn-taking system. From the point of 
view of the speaker on the other hand, greater-than-dyadic conversations present special 
problems: first that of distinguishing between all-other directed utterances and 
specific-other directed utterances, and second that of maintaining the floor in a 
specific-other directed utterance against competition from 'other' others. Gaze is much 
more effective than verbal behaviour in these (greater-than-dyadic) situations because 
it is so much more highly directional and allows a speaker and selected listener to insulate 
themselves from others. 

" What is possibly an extreme version of this is Telefol (New Guinea) 'indirect speech' 
(D. Jorgensen, pers. comm.) where "your friend is hungry" or "your friend departs" are 
found in place of "I'm hungry" and "I'm leaving." I do not know if this can be related to 
the off-noted Melanesian assertiveness (cf. Schieffelin 1975: 117-134). Occupying another's 
place (symbolically invading his body) is certainly open to that sort of interpretation. 

' There is a further exploration to be made of interactional style and style of 
argumentation which requires a short detour to New Guinea. The cover of the third volume 
of Margaret Mead's The Mountain Arapesh (1971) shows two Arapesh standing face to 


face blowing flutes. Williams (1930, cited in Schwimmer 1973:183) reports that in Orokaiva 
sacred flutes which are used in initiation ceremonies are played in pairs by two men facing 
one another. The flutes are played antiphonally (Schwimmer 1973:183), i.e. conversationally. 
The Orokaiva and the Arapesh have societies which are characterized by moiety divisions, 
intra-village feasting, village endogamy and restricted marriage exchange systems. There 
are other groups in New Guinea, however, which have exogamous communities, generalized 
marriage exchange systems, and inter-village feasting with chain-like sequences of 
ceremonial exchange. In at least two of these, the Gimi (Gillison 1977) and the Naugla 
(Nilles 1940), initiation flutes are played side-by-side and not in antiphonal fashion. 

In the East African cultures 1 have studied flutes are not used ceremonially as far 
as I know, but singing is very clearly not antiphonal: it is responsorial and choral. Marriage 
systems are complex, moieties non-existent, etc., but what 1 find most striking about 
this comparison is the implication for political order that seems to be part of the 
face-to-face versus the side-by-side. The Orokaiva and Arapesh, etc. are characterized 
by 'big-men' — a tendency to crystallize into opposing groups around leaders. With the 
Kipsigis and other African groups, one doesn't find this at all, but rather a system which 
is egalitarian in conception and gives to each member of the community the opportunity 
to be heard out and to participate in a consensual decision (Peristiany 1939). 

8. This particular version is from Knapp (1972:12) who gives no specific reference. 

9. This paper has been concerned with one kind of body movement, gaze, in one kind of 
context, conversation. It would not only be unwarranted, but it would be incorrect to 
generalize beyond these limits. In East Africa, Africans can (and in fact do so as a normal 
acitivity) identify individuals ethnically by gait, posture and other aspects of body movement 
and positioning. In addition there is a repertoire of gestures. These gestures often have 
lexical equivalents and may accompany such equivalents when uttered in conversation. 
Body movements also appear to have some function in regulating turn-taking and in 
preventing overlap. The last two types and functions of body movements are studied in 
Creider 1977 and 1978 respectively. The first type, social categorization, has not been 
studied, but it would be highly interesting to do so. 


BECKER, A.L. 1979. The figure a sentence makes: an interpretation of a Classical 

Malay sentence. In T. Givon, ed., Discourse and syntax, pp. 243-259. New York: 

Academic Press. 
BIRDWHISTELL, R. 1971. Kinesics and context. Philadelphia: University of Pennsylvania 

CREIDER, C. 1977. Quantitative aspects of conversational interaction. In D. Sankoff, ed., 

Linguistic variation: models and methods. New York: Academic Press. 
CREIDER, C. 1977. Towards a description of East African gestures. Sign Language Studies 

EKMAN, P. and W.V. Friesen. 1974. Detecting deception from the body or face. Journal 

of Personality and Social Psychology 29.288-298. 
GOODWIN, Charles. 1981. Conversational organization. Interaction between speakers 

and hearers. New York: Academic Press. 
GREENBERG, J.H. 1960. A quantitative approach to the morphological typology of 

language. International Journal of American Linguistics 26.178-194. 
HALLIDAY, M.A.K. and R. Hasan. 1976. Cohesion in English. London: Longmans. 
KEENAN, E.L. 1976. The logical diversity of natural languages. Annals of the New York 

Academy of Sciences 28.73-92. 


KENDON, A. 1970. Movement coordination in social interaction: some examples described. 

Acta Psychologica 32.100-125. 
KIMBALL, J. 1968. On facing. Bloomington: Indiana University Linguistics Club. 
KNAPP, M.L. 1972. Nonverbal communication in human interaction. New York: Holt, 

Rinehart and Winston. 
MOERMAN, M. 1977. The preference for self-correction in a Tai conversational corpus. 

Language 53.872-882. 
PERISTIANY, J.G. 1939. The social institutions of the Kipsigis. London: Routledge 

and Kegan Paul. 
POPE, E. 1973. Question-answering systems. In C. Corum, T.C. Smith-Stark and 

A. Weiser, eds., Papers from the Ninth Regional Meeting, pp. 482-492. Chicago: 

Chicago Linguistic Society. 
SACKS, H. 1972. On some puns with some intimations. Report of the 23rd Annual Round 

Table Meeting on Linguistics and Language Studies. In R.W. Shuy, ed., 

Washington, D.C.: Georgetown University Press. 
SACKS, H., E.A. Schegloff and G. Jefferson". 1974. A simplest systematics for 

the organization of turn-taking for conversation. Language 50.696-735. 
SARLES, H. 1977. After metaphysics: towards a grammar of interaction and discourse. 

Lisse: Peter de Ridder Press. 
SCHEFLEN, A.E. and A. Scheflen. 1972. Body language and the social order. Englewood 

Cliffs, N.J.: Prentice-Hall. 
SCHEGLOFF, E.A., G. Jefferson and H. Sacks. 1977. The preference for self-correction 

in the organization of repair in conversation. Language 53.361-382. 
SCHIEFFELIN, E.L. 1976. The sorrow of the lonely and the burning of the dancers. 

New York: St. Martin's Press. 
WATSON, O.M. 1970. Proxemic behavior: a cross-cultural study. The Hague: Mouton. 

Studies in the Linguistic Sciences 
Volume 14, Number 2, Fall 1984 


Nicholas Faraclas 

This paper proposes an analysis in which Rivers Pidgin 
English (RPE) is basically viewed as a stress-accent language. 
In particular, the correlation of pitch patterns over words 
(when considered in isolation) which have entered RPE from 
stress dominant languages (such as English) with pitch patterns 
found over corresponding words in the acrolect seems to charac- 
terize RPE as a stress-accent language. It is argued here 
that a stress analysis fails to account for the behavior of 
pitch over word groups in connected speech. These words divide 
themselves into phrases that bear one of two basic pitch melo- 
dies which resemble words in many pitch-accent-dominant lan- 
guages. An analysis of pitch in RPE in terms of pitch-accent 
is seen to be inadequate in handling words that have entered 
RPE from some tone-dominant basilectal languages such as 
Yoruba. It is argued here that an analysis which incorporate 
stress-accent, pitch-accent, and tone is necessary in account- 
ing for patterns of pitch-related supra-segmentals found over 
RPE utterances in a comprehensive and yet elegant way. 

1 . Introduction 

1.1 Purpose and scope of study 

This study represents an attempt to analyze all pitch-related 
suprasegmental phenomena which occur over strings of natural speech 
in a language which seems to resist easy classification under the tra- 
ditional rubrics of 'tone language', 'pitch-accent language' or 'stress 
language'. The goal of the analysis is both to describe the patterns 
found in the data in the simplest and most general way without ignoring 
patterns which may not fit very nicely into an extremely abbreviated, 
streamlined interpretation, as well as to reflect the processes (uni- 
versalor language-specific, diachronic or synchronic, pragmatic, morpho- 
syntactic, etc.) which motivate the systems under consideration and 
integrate them into a cognitive superstructure which can function as a 
unitary whole. 

1. 2 Language situation and sociolinguistic history 

Rivers Pidgin English is the dialect of West African Pidgin (or 
Creole, see Hancock 1983) English (WAGE) spoken by at least five million 
people inhabiting the Rivers, Cross River, Imo, and Anambra states of 
southeastern Nigeria, especially in the urban centers such as Port 
Harcourt, Calabar, Aba, Owerri, Enugu, etc. The peoples of southeastern 
Nigeria speak well over 60 clearly differentiated languages (which may 
be subdivided into at least 200 distinct dialects) belonging to no less 
than six separate subbranches of the Niger-Congo family of languages. 


The various ethno- linguistic groups of southeastern Nigeria in 
general, and the Niger Delta in particular have traditionally maintained 
important relationships of exchange at all levels. Such relationships, 
especially those in the economic sphere (trading, etc.) would have re- 
quired the acquisition by members of many different language groups of 
a common language to be used in the marketplace. Indeed, bi- and multi- 
lingualism are the norm rather than the exception in the Delta and such 
languages as Igbo and RPE have been used by Delta peoples as trade lan- 
guages as well as to meet the other communication needs of people of 
diverse linguistic origins living together. 

In the Delta, contact with traders from Europe has been long (dating 
from the 15th century) and sustained. Although the Portuguese, the Dutch, 
the French, the Germans, the Danish and others traded in the Delta, the 
English succeeded in establishing their hegemony over the area by the 
middle of the 19th century. Along with British colonialism came European 
education via missionaries, many of whom were Krio speakers from Sierra 
Leone (who were ex-slaves or descendants of 'repatriated' slaves from the 
Caribbean) . 

Whether RPE developed from the marketplace contact situation between 
European (primarily English) traders and traders of the various Delta 
ethnic groups or from the influence of missionaries from Sierra Leone 
remains undetermined. One must be careful not to overemphasize the role 
of either the traders or the missionaries in the evolution of RPE because, 
except at its earliest stages of development, the language has been used 
primarily as a means of communication among Delta peoples rather than 
between Delta peoples and traders, missionaries, or others from outside. 

It is perhaps unfortuante, but in any case very misleading to have 
almost all of the West African piginized, creolized, and decreolized 
speech varieties in which English has played the role of acrolect or 
superstrate referred to as 'Pidgin' or 'Pidgin English' (Krio, luckily, 
was spared this inappropriate title ). The RPE speech community as well 
as those of the languages resembling it (Cameroonian Pidgin English, 
other varieties of Nigerian Pidgin English, etc.) encompass the entire 
pidginization-creolization-decreolization continuum. For example, for 
a market woman from Eleme (in Rivers state) whose use of Pidgin is 
restricted to business transactions, RPE is a pidgin in the true sense 
of the word, but for her child who uses Pidgin with his playmates from 
Okrika in the market, RPE is a depidginized or creolizing speech form, 
rather than a true pidgin. For the Ikwere man who speaks Pidgin with 
his wife from Nembe and especially for his children, who speak Pidgin 
with their parents, RPE is not a pidgin at all, but a Creole. For the 
child from Port Harcourt who grows up speaking Pidgin, but who hears 
Nigerian Standard English at home (on formal occasions) from his or her 
university educated parents, at school, and on the radio and television, 
RPE is in all probability a decreolized speech form. 

The number of speakers of West African Creole English (RPE included) 
has increased dramatically since the Civil War (1968-1970), and it is 
without a doubt the most widely spoken language in Nigeria at present. 
As is the case with many pidgins and Creoles, WAGE and RPE are spoken 
with varying degrees of similarity to the acrolect (now Nigerian Standard 
English, Mafeni 1971) and/or to the basilects (Igbo, Efik, Ijo, etc.) 
according to both the competence of the individual speaker as well as to 


the sociolinguistic context in which a given utterance occurs (Faraclas 
et al. 1983). The number of people speaking RPE (often alon-rside one or 
more other languages) from infancy is also increasing rapidly. Ihe RPE 
forms used in this work will be those typically used in casual settings 
among speakers who learned the language early in life and who have con- 
tinued to speak it on a daily basis. 

While WAGE is perhaps the most logical choice for a national language 
for Nigeria in terms of the number and geographic distribution of speakers, 
ease of learning for speakers of most Nigerian languages, ethnic 'neutality', 
etc., it is almost unanimously considered by those responsible for the 
formulation and implementation of language policy in Nigeria to be totally 
unacceptable. Most Nigerians, however, including those with a perfect com- 
mand of Standard English, prefer to use Pidgin instead of Standard English 
in interethnic contexts on all but the most formal occasions. Pidgin is 
the language used, for example, among university students from different 
linguistic backgrounds outside of class. Pidgin is clearly the language 
of solidarity in Nigerian society, but Standard English is the language 
of prestige. 

IVhile the full exploitation of Pidgin as a means of mass communication 
has not gone nearly as far in Nigeria as in other nations where the autho- 
rities have adopted a more scientific and rational attitude towards cre- 
olized speech forms (such as Papua New Guinea) , some important steps have 
nonetheless been taken at the local level to increase the use of Pidgin in 
the media. Radio Rivers Two, the state controlled radio station in Port 
Harcourt, now presents public service messages, commercial messages, and 
newscasts in 'Special English' (RPE), while several very popular soap- 
opera type series in WAGE have run on state controlled television stations. 

2.0 Pitch patterns in RPE 

In Rivers Pidgin English we find such minimal pairs of words as (an 
accute accent indicates high pitch, a grave accent indicates low pitch, 
and an apostrophe precedes a stressed syllable): 

a. /go/ '(to) go' [a go 'make t ] 'I went to market.' 

b. /gh/ future marker [a go go ^nakfet 'I will go to market.' 

c. /moda/ 'mother (biological)' ['moda] 

d. /moda/ 'school marm' ['mod^] 

Is the pitch distinction between the two members of each pair cited above 
best analyzed as tonal (as in Gokana ba 'hand' vs, ba_ 'eat'), stress- 
related (as in Englisn 'import vs. in^ort ) , or part of a pitch-accent system 
(as in I jo aklf 'tooth' vs. aka 'maize')? Some of the arguments for and 
against each analysis are presented below. 

2.1 Pitch as stress 

When RPE words are considered in isolation , it becomes apparent that 
the great majority (perhaps 85% - 90T) of words brought into the language 
from languages with stress-dominant pitch patterns (i.e. Portuguese, English, 
Spanish, etc.) bear a high (or, if word-final, falling) pitch over the 
syllable which bears stress in the source language and carry low pitch over 
the other syllables. Since most RPE words are English-derived, most of 
them exhibit the pitch pattern just described. For expository purposes, 


such words will be called Group A (stress-source) words in this work. 
Examples (falling pitch is symbolized by ^. Orthographic representations 
are those recommended by Faraclas et al. 1983): 

a. fada [ f ^da ] 'father' d. latrin [ latrtn] 'latrine' 

b. anoda [anoda] 'another' e. pikin [piktn] 'child' (Portuguese) 

c. parabul [ parabu I ] 'parable' f. panya [papa] 'Spanish' (Spanish) 

The pitch patterns described for Group A words seem to reflect productive 
processes in RPE. Consider the following items of recent origin: 

a. kondokta [kondokta] 'conductor' 

b. drayva [drajva] 'driver' 

c. pitakwa [pltak a] 'Port Harcourt' 

d. jagbajantis [d^aqbad^ant^s] 'junk' (Ijo + English?) 

e. ngwongwobiliti [ pg opg obflUl] ' likeabiiity' (Igbo + English) 

Another interesting parallel between stress patterns in English and 
pitch patterns in RPE is the fact that words with a greater grammatical than 
lexical function which are normally unstressed in English (i.e., non-focussed 
subject pronouns, auxiliary verbs, prepositions, etc.) normally bear low 
pitch in RPE. 


a. go^ [go] future auxiliary vs, £0 'to go' 

b. fo [fo] preposition vs. fJ 'four' 

c. dem [dem] 'they' (non-focussed) vs. dim 'they' (focussed) 

In many (but not all) stress languages, stress is signalled by 
increased length and/or amplitude as well as by high or gliding pitch. 
In RPE, high or gliding pitch are the only reliable cues marking the 
syllables of Group A words which correspond to stressed syllables in 
their source- language cognates. The exclusive use of pitch to signal 
accentuation is more typical of pitch-accent languages than of stress 

2.2 Pitch as accent 

When not in isolation, the above outlined stress-like pitch patterns 
over Group A words occur only at the end of what will be called pitch phrases 
in this work. Pitch phrases in RPE are identical to the phrases over which 
pitch-accent is assigned in such languages as Japanese (McCawley 1965) or 
Jjo (Williamson 1966, Efere 1981). A pitch phrase normally consists of 
(11 symbolizes a pitch phrase boundary): 


i) II wi go kari yam I C || w^ go karf jam \\ ] 
'We will carry yams.* 


ii) II wi go kari yam )| fo tawn j| [tt wl go k^rf jamj/ fo tawn || ] 
'We will carry yams in town.' 



iii) 11 ''i|jwi go kari yam [I fo tawnll [ jj wj_l/ w"i go karf Jam/|fo tawn|/ ] 
'As for us, we will carry yams in town.' 

iv) ]| wi go kariJl yani|| fo tawn // [ j( w"! go karl )) jamj/fo tawnll ] 
'We will carry yams (not cassava) in town. ' 

Notice that in the above examples, the verb kari bears two high pitches when 
not in pitch phrase-final position. Only when kari occurs at the end of a 
pitch phrase (or in isolation) does it bear the high-low pitch pattern which 
corresponds to the stress pattern over the verb carry in English. All Group 
A words carry high pitch over all syllables following the syllable which 
corresponds to the stressed syllable in the source language, unless they 
occur in pitch-phrase final position. Word final high pitches do not fa 1 
unless they are pitch phrase-final as well. 


■ a. anoda [ anoda || ] 


b. wan [ wan 11 ] 

'one' , ^ ^ , , . II 

c. giv mi anoda wan [ giv m\ anoda wan j| ] 

'Give me another one.' 

d. dem de layk wan anoda [ dem de lajk wan anoda |( ] 

'They like each other. ' 
piki n [ P^ Kfn || ] 

gud [ gud||] 

gud pikin [ gud plktn || ] 

'good children' 
pikin gud [ pi kfn gud || ] 

'Children are good.' 

The 10 - 15% of words brought from stress-dominant languages into RPE 
whose pitch patterns do not correspond to those found over their counter- 
parts in the source langauge (even preceding pitch phrase boundaries) di- 
vides into two groups, which will be called Group B words and Group C words 
in this work. Group B words simply carry more than one high pitch. 


a) wuman [ wuman |/ ] 'woman' 

b) an i ma I [ an 'ma I || ] 'animal' 

c) stanop [ stanop II ] 'stand' 

d) mochwar i [motjwarl || ] 'ice-fish' 

L ]) dl wum^n stanop ^ ^op m(5tjw^rf supg ] 'The woman stood eating 

ice-fish soup. ' 

Since the membership of a given word in Group A vs. Group B cannot be 
predicted, some system of marking the distinction between the two groups is nec- 
essary in phonemic representations. The need to mark distinctions between 
stress-dominant source language words with differing pitch patterns becomes 
imperative when Group C words are considered. 





[ moda ] 'mother' vs. 
[ sfsta ] 'sister' vs. 

GROUP C: [ moda' ] 'school mother' 
GROUP C: [ sist^ ] 'nurse' 

Non pitch phrase-final Group C words carry low pitch over all syllables. 
In pitch phrase-final position, the final syllable of a Group C word bears 
high pitch which never falls. Monosyllabic words of this group bear rising 



wot a 

sabi [ 

kruman sabi wota [ 

wota sabi ki i pesin [ 

[ wota ] 


a b f ] ' know ' 

kruman sab* wota|/] 'Sailors know the water.' 
wota sabf y kfl p£sm|/ ] 'Water can kill you.' 

Poser (1984) cites the fact that Japanese exhibits a limited number of 
pitch patterns compared to the possible number of pitch patterns (given the 
basic levels of pitch realized over utterances) as the primary criterion 
for classifying it as a pitch-accent system, rather than as a tonal system. 
Similar criteria are defined by Kingston (1983) for Bantu pitch systems. 
As shown above, words in RPE from stress -dominant source languages bear only 
a small number of the pitch patterns which might occur over them, even in a 
system with only two distinctive levels of pitch. In Ijo (Williamson 1966, 
Efere 1981) four classes of nouns exist, each having a different pitch melody 
associated with it. A pitch-accent account for the behavior of stress-source 
words in RPE might posit the following classes and rules: 


CLASS 1 - 
CLASS 2 - 
CLASS 3 - 


a nod a [ anoda H ] 
pi kl^n [ pfktn f ] 
wota [ wota |j ] 

' another ' 



RULE 1: 
RULE 2: 

RULE 3: 
RULE 4: 

RULE 5: 





WORD (CLASSES 1 and 2) . 





Sample derivations based on the above rules: 

CL. 1 
COND. 1 
COND. 2 
RULE 1: 
RULE 2: 
RULE 3: 
RULE 4: 
RULE 5: 



)(/ CL. 2: 

anoda 11 

pi kin 
pi kfn 

i kTnJ 

pi ktnH 

[pikfn] [pikinll] 

CL. 3: 






[wota] [wot^O] 


While the pitch-accent system outlined above accounts for almost all 
occurrences of words from stress-dominant languages, it does not account 
for certain reduplicated forms, where a rising pitch melody spreads over 
the entire word. 


a) waka [ waka|J] 'walk' wakawaka [ wakawaka||] 'constant moving about' 

b) ha la [ h^lajj] 'yell' ha I ana I a [ halahalajj] 'constant yelling' 

More importantly, however, most of the words brought into RPE from non- 
stress-dominant languages (i.e., languages with strong tonal or pitch- 
accent systems) cannot be accounted for by the pitch-accent system posited 
for stress-words above. 

2.3 Pitch as tone 

Class D words . Class D or non-stress dominant source language words 
may bear almost any possible combination of two level pitches [high and 
low) and are not sensitive to the accent- or stress-like pitch phenomena 
associated with pitch phrase boundaries which have been outlined above. 
There is therefore no logical reason not to analyze pitch patterns occurring 
over Class D words tonally. 

akom gjj^ [||akom||]< IGBO 'fever' 

akom shotop [||akDni JotSplJ] 'malaria medicine' 

b) loy loy invariably [//lojloj//] (YORUBA?) 'specially pounded cassava' 

c) ajagarana - okpokpo invariably [ad^agarana bkpokpoj/]^ IJO 'junk' 

As shown in section 2.2, however, full specification of pitch over 
every syllable is not necessary underlyingly for most words in RPE. 

3. RPE as a mixed tone-accent-stress language 

There are clearly two sets of lexical items in RPE; one set which is 
underspecif ied for pitch underlyingly (Classes A, B, and C) and which allows 
phrase-level accentuation to partially determine surface pitch patterns, and 
another set (Class D) which is fully specified for pitch underlyingly and 
is not affected by phrase-level accentuation. What is the simplest account 
we can give of the behavior of both of these sets of words with respect to 
pitch which captures all of their similarities while ignoring none of 
their differences? 

3. 1 Evidence in support of a mixed system 

Since non-stress-source items need to be fully specified for tone as 
part of a two level tonal system, we must posit two tonemes, high ('j and 
low (y) . These tonemes may be used, however, to account for the behavior 
of stress-source words in a simpler yet more adquate way than the pitch- 
accent system proposed in section 2.2 above, using the following conditions 
and rules; which apply to forms underspecified for tone underlyingly: 


COND. 1: 

COND. 2: 

RULE 1: 
RULE 2: 

RULE 3: 

RULE 4: 





H h"^'^ H H H H'^*^ L _ . 

COND. 1 § 2 /anoda/ /anDda|(/ /animil/ /animl^l || / /wota/ /wota || / 

iTl H HL Hi 

— anoda^/ anifnal|/ — wotaU 



anid^ animal 

\ H ^ L H L 

anoda anidajl 

H.. i^ 




SURFACE FORM [anoda] [anodafl] [anfm^l] [anfmalfj] [wota] [wot^|f] 

Note that this account also handles reduplicated forms (which could not be 
dealt with using the pitch-accent model; see 2.2 above) if they are assigned 
a single low tone. 


COND. 1 § 2 




L H 

L L H H 


3.2 Conclusions and theoretical implications 

The only satisfactory solution to the problem posed by pitch in RPE, 
that is, the only analysis that can predict pitch patterns over utterances 
in a unified way involves the interaction of tonal, pitch-accent, and stress 
units. Tone is assigned to words lexically, words from tonal languages being 
fully specified (one tone per syllable) and words from English being under- 
specified in most cases (often one tone per word) . Underspecif ied items would 


then be assigned additional pitches on the basis of their position within 
a stress-accent group or phrase as well as in relation to the type of 
tone assigned to them lexically. The existence of a mixed system of pitch 
assignment and realization in RPE reflects the mixed origins of the language 
which include stress languages (English, Portuguese) tonal languages (Igbo, 
Yoruba, etc.) and pitch-accent languages (I jo). At the surface, one is 
tempted to apply a stress analysis but the actual system is in many ways 
a reinterpretation of stress in terms of tone and pitch-accent. (A parallel 
case of reinterpretation (in the opposite direction) has been described by 
Li (1984) for Baonan, a Sino-Turkic Creole spoken in north-central China.) 

The RPE data re-affirms the importance of substrate-basilectal speech 
patterns on the underlying structures and processes in pidgins and Creoles. 
In also indicates that speakers of pidgins and Creoles operate from unified 
systems rather than from a loosely bound set of parasystems, each correspon- 
ding to the system as it exists in one or another of the input languages. 
In other words, a pidgin or a creole may behave at the surface in a way 
which is very much like the acrolect and perhaps very much unlike the 
basilect. A more careful analysis usually results in the discovery that 
the strategies used by speakers in the production and processing of surface 
forms are strikingly similar to those typical of basilect, rather than 
acrolect speakers. 

Linguists are only now coming to recognize the importance of pro- 
cesses such as pidginization/creolization in the devlopment of all lan- 
guages. One of the implications of this realization will have to be the 
recognition of the possibility that mixed systems like the one outlined 
for RPE above are not restricted to languages generally classified as 
pidgins or Creoles but may in fact be quite widespread and that such 
mixed systems, where they do exist, function as unified systems rather 
than as sets of parasystems. This may explain the recent successes of 
the autosegmental model, which in effect is supple enough to at least 
begin to accommodate itself to the analysis of systems which have both 
tonal and accentual characteristics. 

There seems to have been a certain reluctance on the part of many 
Africanists to even consider the existence of any non-tonal pitch phenom- 
ena in the languages that they study. It is obvious that such attitudes 
can do nothing to advance, but much to hinder the scientific description 
and analysis of African languages. In many parts of Africa intermarriage, 
hi- and multilingualism, and other forms of interethnic contact and exchange 
are the rule rather than the exception. The traditional bias in linguistics 
against pidginization/creolization as a model for language change has had 
for this reason even more of a negative impact on the study of African lan- 
guages than it has had in other areas. Africanists must learn to recognize, 
describe, and analyze mixed systems wherever they occur. Mixed systems can 
in fact shed much light on some unresolved questions facing the linguist 
in Africa. For example, the analysis of RPE pitch patterns as the result 
of a mixed system has allowed the division of lexical items into distinct 
classes, each of which corresponds to a particular historical state or 
source of borrowing of vocabulary items into the language. Class A words 
appear to have been brought into RPE directly from British and, later, 
Nigerian dialects of Standard English. Class C words reflect pitch patterns 
commonly found in Sierra Leonian Krio (Fyle and Jones 1980) and were probably 
introduced by Krio speakers during the late 19th century. Class B words 


often seem to be the result of an old compounding process which has been 
replaced by the still productive Class A compounding pattern (as in 2.1 
above) on one hand, and by low-toned reduplication (see Section 3.1 above) 
on the other. Class D words are clearly borrowings from other Nigerian 
languages and appear to be increasing in number along with the rise of 
nationalism in the post-colonial era. 

In comparison with some of the neighboring dialects of West African 
Pidgin English (especially the Benin City-Warri-Sapele dialect to the west 
and the Cameroonian dialects to the east). Rivers Pidgin English appears 
to have undergone a limited but nonetheless significant process of decre- 
olization. For example, many words which have final consonant clusters 
in RPE and Standard English have single final consonants in other dialects. 
Voiced plosives also occur word finally in RPE as in Standard English, 
where voiceless plosives occur in neighboring Pidgin speaking communities. 
In general, however, RPE phonology does not differ much from what has been 
reported for other Nigerian and Cameroonian dialects of West African Pidgin 


EFERE, E. 1981. Pitch-accent in the Bumo and Nembe dialects of IJ9: 
a comparison. B.A. Long essay. University of Port Harcourt. 

FARACLAS, N., 0. Ibim, G. Worukwo, A. Minah, and A. Tariah. 1983. A 
language synopsis of Rivers Pidgin English. Language Synopsis 1, 
Department of Linguistics and African Languages, University of 
Port Harcourt. 

FYLE, C. and E. Jones. 1980. A Krio-English dictionary. London: Oxford 
University Press and Sierra Leone University Press. 

HANCOCK, I. 1982. Aspects of standardization in West African Creole 
English. In Proceedings of the 1982 Mid-America Linguistics 
Conference. Lawrence: Kansas University Press, 294-307. 

KINGSTON, J. 1983. The tone-accent transition in Bantu. Manuscript, 
University of California, Berkeley. 

LI, C. 1984. Baonan: Language in contact and language change. Collo- 
quium, Department of Linguistics, University of California, Berkeley. 
March, 1984. 

MAFENI, B. 1971. Nigerian Pidgin. In J. Spencer ed.: The English 
language in West Africa. London: Longman. 

POSER, W. 1984. Downstep and declination in Japanese. Colloquium, 
Department of Linguistics, University of California, Berkeley. 
March, 1984. 

WILLIAMSON, K. 1966. Pitch and accent in Ijo. Linguistics Colloquium, 
Massachusetts Institute of Technology. 

Studies in the Linguistic Sciences 
Volume 14, Number 2, Fall 1984 


Hussein Ali Obeidat 

The phenomenon of relative clause formation (RCF) 
in Arabic, as discussed by Anshen and Schreiber (1968), 
Lewkowicz (1971), and Killean (1972) will be reexamined. 
This paper will argue that (1) RCF in Arabic is a movement 
rule which consists of: (A) relative pronominalizat ion of 
the NP in the embedded clause, and (B) relative pronominal 
fronting into the head position of the relative clause; (2) 
Case assignment is a post cyclic rule; and (3) Grammatical 
agreement in Arabic is a crucial phenomenon in accounting 
for language facts. In conclusion, the paper will briefly 
discuss the theoretical implications of the study and sug- 
gest other types of relative clauses for further investi- 


It has been suggested in three recent studies (Anshen & Schreiber 1968; 
Lewkowicz 1971; Killean 1972) in the transformational literature that rela- 
tive clauses in Standard Arabic (hereafter Arabic) are uncomplicated struc- 
tures that can be derived from topic-comment sentences by applying some sort 
of a copying-deletion rule. These studies reject a movement and pronominal- 
ization analysis of relative clauses in this language by arguing essentially 
that Arabic does not have true relative pronouns, but uses instead a so- 
called "special form of the definite article" (Lewkowicz 197 1:819). This 
special form of the definite article is alladi . We will attempt to show 
here that these analyses are erroneous. 

The objective of this study is to propose an alternative analysis of 
relative clause in Arabic that will provide a better characterization of the 
phenomenon of relat ivizat ion in the language. It will be shown here that 
the above mentioned studies were based on an erroneous analysis of Arabic 
basic sentence structure and on a limited sample of relative clause con- 
structions. This paper examines several types of relative clauses, and pro- 
poses an analysis within the framework of EST that views relativization in 
this language as involving two basic rules: relative pronominalization and 
NP-movement. Several arguments are given in support of this analysis, and 
counter-examples are presented against the topic-comment approach. 


Arabic has relative clauses such as (1) and (2) below where the former 
exemplifies subject relativization and the latter object relativization: 


(1) ?al-waladu alladi maata abuhu. 

the boy who died father his 
(The boy whose father died.) 


(2) ?at-ta9aamu alladi ?akala-hu al-waladu. 

the food which ate it the boy 
(The food which the boy ate.) 

Two main approaches have been suggested in several studies to account for 
data like (1) and (2). These are the "topic-comment" analysis and the "de- 
letion" analysis. In this section, I present an overview of these studies 
and discuss their weaknesses. 

2.1 Topic-Comment Approach . The first publication in transformational 
grammar to deal with RCF in Arabic was Anshen and Schreiber (1968). Anshen 
and Schreiber assume that Arabic is a SVO language. Their assumption is 
based on evidence from verbal agreement. The authors state that "any noun 
in a sentence, other than the first member of a construct phrase, may op- 
tionally be reproduced at the beginning of the sentence" (Anshen & Schreiber 
1968:795). The occurrence of such a noun at a sentence initial position is 
accounted for by the following "focus-transformation" rule which is option- 

(3) X - NP - Y -*■ NP' - X - NP - Y 

where NP = NP' and NP is not the first 
member of the construct phrase 

Consider the following examples from their study (Anshen & Schreiber 1968: 

(A) zaara ?al-binta ar-rajulu. 

visited the girl the man 
(The man visited the girl.) 

(5) ?al-binta zaara-haa arrajulu. 

the-girl visited-her the man 
(The man visited the girl.) 

Sentence (4) above is the result of the application of the T. focus rule 
which copied the noun ?albinta to the beginning of the sentence where pro- 
nominalization applied on the basis of coreferentiality . 

Anshen and Schreiber also claim that the agreement on the verb in 
Arabic is not number agreement, but rather, a form which the pronoun in the 
nominative case takes when it is adjacent to the verb. This claim is used 
to argue for the existence of the optionally reproduced noun as in (5) above. 
This analysis, in their view, simplifies the formation of relative clauses. 

A second study, using essentially the same analysis, appeared in 197 1. 
In this paper, Lewkowicz suggests that relative clauses in Arabic are "de- 


rived from underlying comment which could be assigned the structure NP#S#, 
resulting in the comment within comment structure" (Lewkowicz 1971:815). 
To clarify this, I will use the same example that Lewkowicz gave in her 
paper : 

(6) walad-un maata abu-hu. 

boy a died father-his 
(A boy's father died.) 

(7) al-waladu maata abu-al-waladi . 

the-boy died father-the-boy 
(The boy's father died.) 

According to Lewkowicz, sentence (.7) is first embedded in sentence (6), then 
undergoes the process of relativization producing sentence (8): 

(8) al-waladu alladi maata abu-hu. 

the-boy who died father-his 
(The boy whose father died.) 

Lewkowicz also claims that the relative pronoun alladi is inserted before 
the relative clause "though this insertion might be better handled by re- 
garding alladi as a special form of the definite article" (197 1:819). The 
author does not give any indication as to the source of the relative pro- 
noun. Her argument draws only on the occurrence of the relative pronoun 
alladi after a definite NP. 

About a year later, Killean (1972) published a paper in which she sup- 
ports Lewkowicz 's analysis. According to Killean, the order of relative 
clause formation processes is as follows: 

"(1) Topic-comment extraposition in SI with simultaneous re- 
placive pronoun formation or, if not simultaneous, at least 
before the boundaries are erased. (2) Embedding of this 
topic-comment into a higher NP . (3) If the embedding of 2 
is well formed, i.e. if the two left-most NP nodes are non 
distinct, the formation of a relative pronoun out of the 
definite article L^ plus - adi , - ati , or, - adiina (based on 
gender-number concord) takes place. (4) The assignment of 
case at the sentence level follows this last transformation" 
( 1972: 148). 

2.2 Deletion Analysis . While Lewkowicz accepts the topic-comment 
analysis mutually proposed by Anshen and Schreiber, her analysis differs 
from theirs in important respects. For example, she assumes a copying- 
deletion rule as presented in (9) below. According to this rule, relative 
clauses are formed by the deletion of a coreferential NP and the insertion 
of a form of the relative pronoun. 


(9.) S NP. r NP - X - VP - Y 

1 2 3 4 5->-iwh0345 

where NPi = NP2 

The analysis proposed by Lewkowicz was accepted by Killean (1972), however, 
Killean was critical of the data Lewkowicz uses as being rare or barely ac- 
ceptable by native speakers of Arabic. Other studies on Arabic RCF (Awwad 
1973; Suaieh 1980) have partially accepted the analysis of Lewkowicz (197 1). 
According to Awwad, the comment within comment approach proposed by Lewkowicz 
has failed to account for the source of the relative pronoun. Hence, Awwad 
proposes a modified rule of RCF in Arabic as shown in (10): 


-[_ Np[ X-NP-YJ J- 

12 4 # 3 4 5 # 6 



Condition 2=4 

2.3. Problems . The analyses summarized above are inadequate in many re- 
spects and fail to account for many relative clause constructions in Arabic. 
Let us consider these inadequacies or problems briefly in turn. 

First, the T. focus transformation proposed by Anshen and Schreiber 
(1968) permits only the first NP, "which is not the first member of a con- 
struct phrase" (1968:795), to be preposed or moved to sentence initial posi- 
tion. Thus, in a sentence where the order is VSO and the NP to be relativ- 
ized is the object, the T. focus rule will reproduce the subject NP which 
will require an adhoc rule to delete it, or, the formulation of another rule 
which will change the order of relations within the phrase so that the NP 
to be relativized will be promoted into the position of the "reproducable" 
NP. The T. focus transformation also will not account for relativization in 
oblique positions since the relativized NP is part of a PP, and Arabic al- 
lows such relativization to take place as will be shown later (ex. section 

Second, the hypothesis that the structure underlying relative clauses 
in Arabic is a topic-comment structure (Lewkowicz 1971; Killean 1972) is 
factually erroneous, because Lewkowicz has based her analysis on sentences 
which are ungrammatical and cannot be generated by Arabic grammar. Further, 
the comment within comment analysis where the embedded comment is the source 
for relativization is equally erroneous. Such an analysis will generate a 
new topic as in (lie) below which will somehow be deleted later in the anal- 
ysis without any indication that it is recoverable. Consider, for example, 
sentence (lid) in relation to (lla-c), as given in Lewkowicz (1971:820): 


(11) a. walad-un maata abu-hu. 

boy a died father-his 
(A boy whose father died.) 

b. al-walad-u maata abu-al-walad-i . 

the-boy died father-the-boy 
(The boy whose father died.) 

c. waladun al-waladu maata abu-hu. 

boy a the-boy died father-his 
(A boy whose father died.) 

d. alwaladu alladi maata abu-hu. 

the boy who died father-his 
(The boy whose father died.) 

According to Lewkowicz, sentence (lib) is embedded in (ila); then it under- 
goes the process of relativizat ion to produce (lie) where there are two 
topics for the sentence: waladun , alwaladu . The topic of (lie) then is 
deleted by some kind of a rule according to Lewkowicz (197 1:820), result- 
ing in (lid) above. This analysis complicates unnecessarily the grammar of 

Third, the topic-comment analysis fails to account for the source of 
the relative pronoun alladi . The morpheme alladi cannot be considered a 
special form of the definite article as Lewkowicz (197 1) claims, or, a var- 
iant of the definite article as Killean (1972) suggests, because it is not 
subst itutable with the definite article a_l, as the ungrammaticality of sen- 
tence (12b) attests: 

(12) a. jaa?a al-waziiru al-sirriiru. 

came the-minister the-evil 
(The evil minister came.) 

b. *jaa?a al-waziiru alladi sirriiru 

came the-minister who evil 

The article al^ is not a reduced form of alladi , but rather, it is a marker 
of def initeness . Sentence (12b) requires the presence of the independent 
resumptive pronoun huwa (he) to be gramnatical as in: 

c. jaa?a al-waziiru alladi huwa sirriiru 
came the-minister who he evil 

The failure of Lewkowicz 's and Killean's accounts of the relative pronoun 
alladi has been correctly observed by Awwad (1973). This author argues 
that alladi is an NP in the embedded sentence being copied to the front po- 
sition of the relative clause, and on the basis of coreferent iality with the 
head, it is realized as a relative pronoun. 


Fourthly and finally, while the copying and pronominal izat ion analysis 
correctly characterizes alladi as a relative pronoun resulting from rule (9) 
or (10), it also fails to explain one important property of relative clauses 
in this language: grammatical agreement. For example, the relative pronoun 
alladi takes case marking depending on the relativized NP. Consider the 
following examples: 

(13) a. a 1 -muadda f - aan i allad- aani qaabalahumaa al-mudiir tarakaa. 

the employee two who two met-them the director left 
nom nom 

(The two employees who the director met left.) 

b. qaabal-tu al-?ustaad- ayni allad- ayni zaar-aa al-jami9ah. 

met-I the teachers two who two visited the university 
ace ace 

(I met the teachers who visited the university.) 

Notice that the relative pronoun alladi is inflected for the nominative in 
(13a) and for the accusative in (13b), whereas the relativized site is quite 
the opposite in both 13 (a) and (b) as exemplified in (c) and (d), respec- 

al-muaddaf- aani 

the-employee Itwo' 

qaabala al-mudiiru al-muaddaf- ayni tarkaa 

the-director the-employee Ftwo left 

qaabal-tu al-?ustaad- ayni 

zaara al ?ustaad- aani al-jami9ah 
visited the teacher 


It is obvious that grammatical agreement changes according to the relativi- 
zed NP. The question to be addressed here is, how can grammatical agreement 
be explained by copying-deletion analysis? Needless to say, there are other 
problems of this type that the copying-deletion analysis failed to account 
for. These problems, as I will argue, can only be resolved by assuming a 
movement type of analysis for Arabic relative clause formation. 


Since the above-discussed approaches have failed to recognize the 
source of relative clauses by adopting a topic-comment type of structure, 
we propose as an alternative that the basic sentence structure in Arabic is 
the source for relative clause formation. 

3.1 The Arabic Sentence Structure . In this paper, I assume that the 
basic sentence types are the sources for RCF. There are two types of sen- 
tence structure in Arabic: (1) Nominal clauses, and (2) Non-nominal clauses. 

This section summarizes the sentence structure of Arabic and other cha- 
racteristics which will facilitate the description of RCF in section (3.3). 


Nominal clauses consist of nominal equational and nominal non-equation- 
al clauses. The former typically include an NP followed by an adjective, 
adverb, or, a preposition phrase. 


ad j . 
adv . 

Examples of this structure are given in (15): 

(15) a. alwaladu fi-d-daari. 

the boy in-the-house 
(The boy is at home) 

b. Mofiammadun qaa?id-un. 

Mofiammadun leader-a 
(Mohammadun is a leader.) 

c. al-waladu hunaak. 

the boy there 
(The boy is there . ) 

Nominal non-equat ional clauses, in contrast, include an NP followed by a VP 
(NP) (PP) (ADV), as exemplified in (16): 

(16) al-waladu dahaba ?ila-l-bayt. 

the-boy went to the-house 
(The boy went home.) 

Non-nominal clauses have been traditionally referred to as "verbal sen- 
tences" by Arab grammarians, and represent the basic word order in Arabic, 

(17) ?akala alwaladu at-t9aama. 

ate the boy the food 
(The boy ate the food.) 

(18) zaara at-tullaabu al-mattiafa al-wataniya. 

visited the-students the museum the national 
(The students visited the national museum.) 

Verbal sentences in Arabic differ from nominal sentences in that they repre- 
sent the unmarked word order. Although the subject in (17) is singular and 
in (18) plural, one could realize that the agreement on the verbs is the 
same. However, this is not the case in nominal non-equational sentences. 
Further discussion is presented in section (3.2). 

3.2 Grammatical Characteristics of Simple Sentences . There are two 
genders in Arabic: masculine and feminine. The former is indicated by zero 
suffix, while the latter by -at. For example: 


(19) a. dahaba al-mudiiru ?ila-l-?ijtimaa9. 

went the director to the meeting 
(The director went to the meeting.) 

b. dahab-at al-mudiiratu ?ila-l-?ijtimaa9 . 

went(f) the director(f) to the meeting 
(The director went to the meeting.) 

Grammatical relations and agreement in Arabic are indicated by morphological 
changes on nouns and verbs as well; as a result, Arabic allows free word or- 
der (Bakir 1980). Perceptual complexity is resolved by case marking. Arabic 
distinguishes three types of cases: nominative, accusative, and genitive. 
Let us examine each of these briefly. 

The nominative case is always assigned to NP in the subject position 
and it is indicated by the final -iJ mode marker, irrespective of the gender. 
The word alwalad-u 'the boy', for example, in (20a) is always recognized as 
a subject because of the nominative case assigned to it. 

(20) a. jaa?a alwalad-tj. 

came the-boy-nom 
(The boy came . ) 


b. zaara al-bint-£ ar-rajul-u. 

visited the-girl the man 
(The man visited the girl.) 

The accusative case in Arabic is indicated by the -a_ suffix marker assigned 
to objects. Notice that the word al-bint-a 'the girl' in (20b) is inflected 
for the accusative, and hence, it is the object. The case marking for the 
genitive is -i; for example: 

c. dahaba ?at-taalib-u ?ila-l-madrasat-i^. 

went the-pupil to the school-gen 
(The pupil went to school.) 

Agreement between the verb and its arguments depends on the word order in 
the sentence. If the ordering is SVO, then the verb must agree with the sub- 
ject in number and gender as exemplified in (21a-d): 

(21) a. ?at-taalib-u dahaba ?ila-l-madrasat-i. 

the pupil-nom went to-the-school-gen 

b. ?at-tullaab-u dahab-uu ?ila-l-madrasati. 

the pupils-nom went-pl to-the-school 

c. ?at-taalibaa-tu dahab-na ?ila-l-madrasat-i. 
the pupils(f) went-pl to the school-gen 


d . ?at-taalib-aani dahab-aa ? ila-1-madrasati 

the pupils-two went-two to the-school 


However, if the verb precedes the expressed subject, then the verb is in the 
singular form, and agreement is required only in gender. For example: 

(22) a. dahab-at at-taalibaatu ?ila-l-madrasati . 

went(f) the-pupils-pl( f ) to-the-school 

b. dahaba at-taalibaani ? ila-1-madrasati . 

went the-pupil-two to-the-school 

The following conclusions can be drawn from the examples in (.20) through 
(22): First, case agreement is a basic property of Arabic. Second, this 
type of agreement is noun governed as should be expected from case marked 
languages. And third, VSO represents the unmarked word order. These pro- 
perties are crucial for any analysis of RCF in Arabic; any analysis that 
fails to take them into account cannot adequately explain the facts of re- 
lativization in this language. 

3.3 The Structure of the Relative Clause . As in many other languages, 
Arabic allows relativization in subject, object, and oblique positions. In 
this section, I consider different types of relative clauses, and discuss 
the interaction of RCF with grammatical agreement. The data examined argue 
against a topic-comment analysis and a copying-deletion analysis. 

3.3.1 Subject Relativization . Subject relativization in Arabic exhi- 
bits a number of syntactic and morphological properties. Consider, for ex- 
ample, the following sentence: 

(23) a. al-walad-u alladi daraba al-bint-a saafara. 

the-boy-nom who hit the-girl-acc left 
(The boy who hit the girl left.) 

This sentence has the structure S + S which consists of the clauses: 

b. ?al-walad-u saafara 
the boy left 

c. daraba al-walad-u al-binta-a 
hit the-boy-nom the-girl-acc 

Clause (b) is a nominal sentence where the NP alwaladu precedes the VP 
saafara, and in this case, agreement is required, as explained in (3.2). 
In Arabic, subject-verb agreement for the third person singular masculine 
is marked by zero suffix; this explains the fact that in both (b) and (c) 
the agreement is zero though each clause has a different word order. Sen- 
tence (23a) is clearly an embedding of both 23 (b) and (c). Notice that 


the subject of (23c) alwaladu , which is embedded in (23b), is no longer 
there, rather we have a relative pronoun materialized in the head position 
of the relative clause. 

The question that comes to mind here is, how can these facts be ana- 
lyzed? Before answering this question, it should be pointed out that the 
relative pronoun in Arabic for the masculine and feminine singular, alladi 
and allati , respectively, is invariable, and hence, we might not be able 
to tell clearly what actually took place. It could be explained as a dele- 
tion of NP in the embedded clause (23c) and insertion of the form alladi , 
or, it could be a relativization of the NP in the embedded clause which is 
realized as a relative pronoun being moved later to the front position in 
the higher clause. Either analysis at this point would account for the 


To select the best analysis, however, additional data must be consid- 

(24) a. hadarat al-bintaani allataani daraba-humaa al-waladu. 

came the girls(two) who hit-them the-boy 
(The girls who(m) the boy hit came.) 

hadarat al-bint-aani 

came the-girl-two 

daraba al-walad-u al-bint-ayni. 


the-boy-nom the-girl-two 

Sentence (24a) consists of two clauses, the main clause and the embedded 
clause, as shown in (24b). Notice that the NP albintayni in the embedded 
clause is the object and is accordingly inflected for the accusative, where- 
as, the NP albintaani in the main clause is the subject and inflected for 
the nominative. After the application of relativization to the structure 
(24c), we get (24d), 


tiadarat al-bint 

daraba al-walad allat. 
pace "I 

hadarat al-bint allat 

Enom"] pnomn 
duall l+duall 

darab alwalad-humaa. 

The relative pronominal in the embedded clause in (24c) is in the accusative 
( allatayni ) , and in (24d) this relative pronominal is assigned a nominative 
case marker allataani. 

How can these morphological changes be accounted for? Evidently, such 
a change can only be explained by suggesting that RCF in Arabic involves a 
movement rule which moves the relative pronoun allataani in (24c) from the 
lower embedded clause to the main clause. It also suggests that case as- 
signment applies after such movement has taken place on the basis of the new 
position it occupies in the sentence resulting in (24a). 


Notice another characteristic of relative clauses exemplified in (24a). 

This sentence contains the pronoun humaa which functions as a resumptive 

pronoun. Its occurrence in (24a) is optional; the relativized site could be 
a gap as in (25) : 

(25) hadarat albintaani allataani daraba-0 alwaladu. 

came the girl(two) who hit the boy 
(The girls who the boy hit came.) 

The realization of the resumptive pronoun is obligatory under the condition 
that deletion of the resumptive pronoun will change the reading of the sen- 
tence as in 26 (a) and (b). 

(26) a. gabal-tu al-walada alladi daraba-hu Ali. 

met-I the boy who hit him Ali 
(I met the boy who(m) Ali hit.) 

b. gabal-tu al-walada alladi daraba Ali. 

roet-I the boy who hit Ali 
(I met the boy who Ali hit.) 

26 (a) and (b) are semantically two distinct sentences. In (26a) the noun 
Ali is the actor, whereas in (26b), Ali is the patient. Deletion of the 
resumptive pronoun in such a case affects the relationship between the verb 
and the arguments . 

It is important to indicate here that in subject relativization in non- 
equational sentences in Arabic, the relativized site (where the relativized 
site is the subject of the embedded clause) is a gap which cannot be filled 
with a resumptive pronoun as in other cases of relativization in other lan- 

3.3.2 Object Relativization . Arabic allows relativization in any ob- 
ject position: direct (D.O.) or indirect (I.O.). As in the case of subject 
relativization, object relativization involves some morphological and syn- 
tactic changes. Consider the sentences in C27) that involve D.O. relativi- 

(27) a. ra?ay-tu al-walad-a. 

saw-I the-boy-acc 
(I saw the boy.) 

b. darabu al-walad-u al-bint-a. 

hit the-boy-nom the-girl-acc 
(The boy hit the girl.) 

That is, given sentence (27a) where al-walada 'the boy' is the object as 
indicated by the accusative case marker and in (27b) where it is the subject 
and inflected for the nominative, relativization in the object position will 
produce (27c): 

c. ra?ay-tu al-walad-a alladi daraba al-binta. 

saw-I the-boy who hit the-girl 
(I saw the boy who hit the girl.) 

What we notice here is that the NP alwalad-u 'the boy' in (27b) is relati- 
vized and a relative pronoun alladi surfaced following the modified NP in 
the matrix clause. The same process holds for sentences 28 (a) through (c) 
below. Note that the NP alwalad-aani 'the two boys' in (28b) differs syn- 
tactically and morphologically from the NP alwalad-ayni in (28a). The 
former is the subject and it is inflected for the nominative, whereas, the 
latter is the object and it is inflected for the accusative. 

(28) a. ra?aytu al-walad-ayni. 

saw-I the boys-two-acc 
(I saw the two boys.) 

b. daraba al-walad-aani al-bint-a. 
hit the-boys-two-nom the-girl-acc 

c. ra?ay-tu al-waladayni alladayni darab-aa al-binta 
saw-I the-boy-two who hit-Ag the girl 

In looking at 28 (a-c) we expect the relative pronominal in (c) to be 
alladaani 'who-nom' . but instead we have alladayni . 

The question which comes to mind here again is, how can these facts 
be analyzed? To answer this question, it could be suggested that RCF in- 
volves two steps: (1) relativization of the NP in the embedded clause, 
alrauhaadirataani 'the two lecturers' in the case of (29b), which is real- 
ized as a relative pronominal allat as in (29c), and (2) a movement of the 
relative pronominal allat from the embedded clause to the matrix clause. 
It follows also that the case assignment rule takes place after the move- 
ment rule has applied, where the relative pronoun agrees in case with the 
preceding NP in the main clause, as shown in (29d), and as a result we get 

(29) a. qaabal-tu al-munaadirat-ayni allat-ayni 

met-I the lecturers(f )-two who-two 

zaara-taa al-jami9ah. 

visited-Ag the-university 

(I met the two lecturers who visited the university.) 


qaabaltu al muhaadir 
+ fem 

al-mutiaadir zaara-taa al-jami9ah. 
+ fem 
+nom _ 


qaabaltu al mutiaadir 
+ fera 

allat zaarataa al jami9ah. 


qaabaltu al mutiaadir allat zaarataa al jami9ah. 

+ fem 


If we were to apply case assignment as a cyclic rule, one would expect the 
relativized NP to move with its nominative case marker allataani 'who' 
rather than allatayni as in (30), and the resulting sentence would be un- 
grammat ical . 

(30) *qaabal-tu al-bintayni allataani zaara-taa al-jami9ah. 

met-I the girl-two who visited-Ag the university 
(I met the two girls who visited the university.) 

In this case, the resumptive pronoun is a gap that cannot be filled, since 
the relativized site in the embedded clause is in the subject position, as 
discussed previously (cf. section 3.3). 

Relativization into indirect object (I.O.) position operates basically 
in the same manner, as can be seen in the examples in (31) through (33): 

(31) a. al-waladu alladi ?a9taa ar-rajulu al-kitaaba la-hu dahaba. 

the boy who gave the-man the book to-him left 
(The boy to whom the man gave the book left.) 

b. al-waladu alladi ?a9taa-hu ar-rajulu al-kitaaba dahaba. 
the-boy who gave-him the-man the-book left 

c. al-waladu alladi ?aitaa- ar-rajulu al-kitaaba dahaba. 
the boy who gave- the man the book left 

(32) ?at-taalibaani alladaani ?ahadaa al-mu9allimu al-jaa?izata 
the-students( two) who awarded the-teacher the-prize 

la-humaa taxarraj-aa. 

to them graduated-Ag 

(The two students to whom the teacher awarded the two prizes 
graduated . ) 

(33) ?al-muhandis-uuna alladiina manatiat-hum al-hukuumatu 
the-engineers-pl who granted-them the-government 

al-jaa?izata taraka. 

the prize left 

(The engineers to whom the government granted the prize left.) 


The examples in 31 (b) and (c) indicate that there is interaction between 
the dative movement rule and relativization. However, Arabic does not re- 
quire I.O. to be promoted to D.O. position to make it more topical and more 
accessible for relativization. The semantic complexity is resolved by case 
marking and the presence of a resumptive pronoun. This holds also for re- 
lativization into oblique positions as well. Kote that the presence of a 
resumptive pronoun is optional in sentence (31) as shown in (31c), where 
the replacive pronoun -hu 'he' does not surface, whereas in both (32) and 
(33) the resumptive pronoun materializes. To illustrate the process of 
I.O. relativization, consider the following: 

(34) a. manatiat ad-dawlatu al-ja?izata li-l-mu9allim-iina . 

granted the-government the-prize to-the-teachers-pl-gen 

b. ?al-mu9allim-uuna tagaa9ad-uu. 

the-teacher-pl-nom retired-Ag 
(The teachers retired.) 

c. ?al-mu9allim-uuna alladiina manatiat ad-dawlatu 
the teacher-pl-nom who granted the government 

al-ja?izata la-hum tagaa9d-uu. 

the-prize to-them retired-Ag 

(The teachers to who(m) the government granted the prize 
retired. ) 

Here again, sentence (34c) is an embedding of (34a) into (b). Notice that 
the head NP al-mu9allimuuna 'the teachers' in (34b) is a nominative subject, 
wheras the same NP in (34a) is an indirect object inflected for the genitive 
- iina because of the preceding preposition. After RCF applies, the NP 
almu9allimiina in the embedded clause relativizes into the head position of 
the relative clause, leaving the resumptive pronoun - hum 'they' behind as 
exemplified in (34a). One of the peculiarities to be noted here is that the 
relative pronoun allad-iina seems not to agree with the head NP al-mu9allim- 
uuna , rather it kept the syntactic features of the relativized NP in the 
embedded clause. To solve this problem, schools of Arabic linguistics have 
adopted two different positions with regard to the form of the plural rela- 
tive pronoun: (1) the plural relative pronoun is invariable regardless of 
case, which is the case in (34c) above, and (2) there are two forms of the 
plural relative pronoun: (a) allad-uuna which is inflected for the nomina- 
tive, and (b) allad-iina which is inflected for the accusative. If we were 
to adopt the latter position, sentence (34c) would be ungrammatical, since 
case marking is suggested to be a final stage rule and the correct form 
would show agreement with the preceding NP in case, number, and gender. 

It should be clear from the examples given thus far, that D.O. and I.O. 
relativization are similar processes. It is only in the case of I.O. rela- 
tivization, where no dative movement has applied, that the retention of a 
resumptive pronoun is obligatory, otherwise, it is optional in both D.O. and 


To summarize sections (3.3.1) and (3.3.2), it has been demonstrated, 
contrary to the topic-comment analysis and deletion analysis (Anshen £< 
Schreiber 1968; Lewkowicz 197 1; Killean 1972; etc.), that RCF in Arabic 
involves a movement rule that consists of: 

A. Relative pronominalizat ion of the NP in 
the embedded clause 

B. Relative pronominal fronting into the 
head position of the relative clause 

It follows that case assignment rule applies finally as a post cyclic rule. 
I will now discuss relativizat ion into oblique positions. 

3.3.3 Relativization into Oblique Position . Keenan and Comrie (1977) 
have proposed that the grammatical function of the NP determines its acces- 
sibility to relativization. As a result, the position of the NP on the AH 
determines the degree of complexity of the relative clause formed on that 
position. Assuming that this proposal is correct, it is predicted that re- 
sumptive pronouns will be used obligatorily when one relativizes in low po- 
sitions on the AH. It will be shown that this prediction is supported by 
the data on Arabic in the discussion hereafter. 

Arabic is one of the languages that permits relativization into obli- 
que positions (oblq hereafter). Consider the following examples: 

(35) ?al-maktabatu allati ?ista9ar-tu al-kitaaba min-haa ba9iidah. 

the-library which borrowed-I the-book from-it far 
(The library from which I borrowed the book is far.) 

(36) ?al-midyatu allati ta9ana al-waladu ar-rajula bi-haa jadiidah. 

the-knife which stabbed the-boy the-man with-it new 
(The knife with which the boy stabbed the man is new.) 

(37) zur-tu al-ma9rada alladi ?istaraa ar-rajulu 
visited-I the-showroom which bought the-man 

as-sayyarata min-hu. 

the-car from-it 

(I visited the showroom from which the man bought the car.) 

The relativized NPs in (35), (36) and (37) are ?al-maktabah 'the library' 
?al-midyah 'the knife', and al-ma9rada 'the showroom' , respectively. All 
of these NPs have been relativized in lower position on the hierarchy as 
indicated by the resumptive pronouns - haa in both (35) and (36) and - hu 
in (37), respectively, which they left behind. Note also that in these 
sentences the relativized NPs are inflected for the genitive as a result 
of being objects of prepositions. However, these NPs after relativization 
has applied are assigned new cases, nominative in (35) and (36), respec- 
tively, because the modified NPs al-maktabat-u and ?al-midyat-u are nomina- 


tive subjects and hence relative pronouns agree with the preceding NP in 
case, number and gender as it has been argued in this paper thus far, where 
case assignment is a post cyclic rule. In (37) the relative pronoun alladi 
is inflected for the accusative as predicted by the case assignment rule 
since the modified NP al-ma9rad-a 'the showroom' is in the object position 
in the matrix clause. However, the case marker on the relative pronoun 
alladi is not transparent. 

Having presented the data on oblq relat ivizat ion, the question that 
arises here is, how can the facts discussed be accounted for? And more 
specifically, what is the process involved in relat ivizat ion into oblq po- 
sition? To explain these facts, the following examples are illustrative: 

(38) a. ?al-ma9rad-aani allad-aani ?istaraa ar-rajulu 
the-showroom-two which-two bought the-man 

as-sayyaraata min-humaa ba9iidaani. 

the-cars from-them far 

(The two showrooms from which the man bought the 
cars are far . ) 


?istaraa ar-rajul as-sayyaaraata 
bought the-man the-cars 

from- the-showroom-two 

l+nom J 



?istaraa ar-rajulu as-sayyaarata 
VP subj-nom obj-acc 

^nom I 

al-ma9rad allad ?istaraa ar-rajulu as-sayyaarata 

ri-dual] pduall 
l+nomj l+nom J 

min-humaa ba9iid 

+RP pduall 


(39) *al-ma9radaani alladayni 
the-showroom-two which 

?istaraa arrajulu assayyaraata 
bought the man the cars 

minhumaa ba9iidaani. 
from-them far 


Relativization of dual and plural nouns in Arabic clearly Indicates the 
morphological and syntactic changes on the relativized NPs according to the 
positions they occupy in sentences. Note in (38b) the NP alma9rad-ayni 
'the two showrooms' in the embedded clause is inflected for the genitive as 
indicated by the suffix - ayni , whereas the relative pronoun allad-aani 
'which' in (38a) is inflected for the nominative as can be noted by the suf- 
fix - aani . Examples (38a-d) explain how such a change came about. As can 
be seen, the NP al-ma9rad-ayni of the embedded clause in (38b) is relativiz- 
ed into the relative pronoun allad as in (38c). This relative pronoun is 
then fronted to the head position of the relative clause as in (38d), leav- 
ing a replacive pronoun - humaa behind. Notice also that the fronted rela- 
tive pronoun allad is assigned a new case marker according to the new posi- 
tion it occupies, which indicates that case assignment is a final stage rule. 
Violation of this rule results in ungrammatical sentences as exemplified in 

To summarize this section, three things have been demonstrated: First, 
relativization into oblq positions involves a movement rule which relati- 
vizes an NP in the embedded clause into a relative pronominal, then this 
relative pronominal is fronted into the head position of the relative clause. 
Second, case assignment applies as a final stage rule, and finally, relati- 
vization into oblq positions in Arabic results in the realization of an ob- 
ligatory resumptive pronoun which brings support to Keenan and Comrie's pro- 
posal indicated at the beginning of this section. 

3.3.4 Relativization and Grammatical Agreement . As may have been no- 
ticed from the discussion thus far, the analysis presented here depends 
heavily on grammatical agreement. We have described the case agreement as- 
pect and its role in the derivation of even simple sentences, but we said 
little about verb agreement. In this section, we focus briefly on both of 
these aspects to demonstrate the validity of our analysis. In this respect, 
consider the following: 

(40) a. dahaba al-walad-u 
went the boy 

b. al-waladu dahaba 
the boy went 

dahabat al-bintu 
went the-girl 

d. al-bint-u dahabat 
the-girl went 

(41) a. dahaba ?al-?awlaadu 
went the-boys 

b. ?al-?awlaadu dahab-uu 
the-boys went-Ag 

c. dahabat al-banaatu 
went the-girls 

d. albanaatu- dahab-na 
the girls went-Ag 

In sentences 40 (a) and (c) the ordering is verb-subject, whereas in 40 (b) 
and (d) the ordering is subject-verb. However, the agreement on the verb 
dahaba(t ) is the same in either order because both nouns alwaladu 'the boy' 
and al-binta 'the girl' are in the singular. In contrast, sentences 41 (b) 
and (d) show different endings than (a) and (c). In the former, the order- 


ing is subject-verb and in this case the verb must agree with the noun in 
number and gender as shown in dahab-uu and dahab-na , respectively, whereas 
in the latter, agreement is shown only in gender as indicated by the suffix 
-t on dahabat in (c). Let us now see how agreement works in relativization 
and what that indicates. Consider the following examples: 

(A2) a. dahaba al-walad-aan ?ila New York. 

went the-boy-two to New York. 
(The two boys went to New York.) 

b. al-walad-aani raja9-aa-. 

the-boy-two came-Ag 
(The two boys came.) 

c. al-waladaani alladaani dahab-aa ?ila New York raja9-aa. 
the-boys-two who went-Ag to New York came-Ag 

Notice that the verb dahaba in (.42a) does not show agreement in number with 
the following subject al-waladaani 'the two boys'. In i>42c) after relati- 
vization applies where (a) is embedded in (b), the verb dahab-aa shows 
agreement with the relative pronoun alladaani as indicated by the suffix 
-aa. These facts demoni:trate that the relativized subject of the embedded 
clause is a dual noun and that the relative pronoun does not agree with the 
preceding head NP but does with the verb in the embedded clause. 

Another aspect that needs to be underscored here is noun morphology 
which is the same as relative pronouns. The inflection for the dual is 
- aani for the nominative and - ayni for the accusative and genitive. The 
inflection for the plurals differs according to the type of plural. In the 
case of 'broken plurals' (which do not follow a defined rule in the way they 
are formed), the inflection is the same as that of the singular: -u_, -a^, 
and -i for the nominative, accusative, and genitive, respectively. In the 
case of 'sound plurals' (which are formed on the basis of defined rules), 
we have to distinguish between masculine sound plurals and feminine sound 
plurals. Masculine sound plurals have the inflection - uuna for the nomina- 
tive, and - iina for the accusative and the genitive; whereas, feminine 
sound plurals have -u for the nominative and -i_ for both accusative and 
genitive . 

Interestingly enough, relative pronouns have the same agreement pat- 
terns as nouns have. Consider this example: 

(43) ra?ay-tu al-mu9allim-iina allad-iina mananat-hum 
saw-I the-teacher-pl wh-Ag granted-them 

addawlatu al-ja?izah. 

the government the-prize 

(I saw the teachers whom the government granted the prize.) 


Note that in (43) the relative pronoun alladiina carries the same agreement 
that the noun almu9a 1 1 ini- i ina has. In the case of masculine and feminine 
singular, the agreement on the relative pronoun, alladi and allati , respec- 
tively, is not transparent. In the case of the dual, the distinction is 
clear and it has the same inflection nouns have, i.e. - aani for nominative 
and - iina for the accusative and genitive. The same is true of plural re- 
lative pronouns (cf. 3.3.2). 

The point here is that if one were to assume that relative pronouns 
are either variants or the definite article - al (cf. Lewkowicz 197 1; Kil- 
lean 1972), or inserted pronouns via a copying-deletion rule (9) and (10), 
it would be impossible to account for all these different agreement mani- 
festations: number, case, and gender. 


This paper has presented an argument for RCF in Arabic. First, it 
has demonstrated that RCF in Arabic is a movement rule and involves the 
following: (a) relative pronominalizat ion of the NP in the embedded 
clause, and (b) relative pronominal fronting into the head position of 
the relative clause. Second, case assignment in Arabic is a post cyclic 
rule. Third, the relative pronoun alladi is a relativized NP in the em- 
bedded clause being fronted into a higher position via a movement rule. 
Fourth, grammatical agreement in Arabic is a crucial phenomenon in all 
attempts to account for language facts. Negligence of this phenomenon 
results in an improper analysis of the language as shown to be the case 
of copying-deletion analysis. Fifth and finally, VSO represents the un- 
marked word order in Arabic. 

Further investigation on relativization in Arabic could concentrate 
on other crucial aspects of relativization such as indirect mode relati- 
vization (i.e. the relativized NP in the embedded clause is syntactically 
different from the corresponding NP in the matrix clause, but yet they 
are semantically understood to be the same). For example: 

(44) -?antumaa alladaani dahab-aa ?ila -al-mubaarati! 

you-dual who went-Ag to the-match 
(You are the ones who went to the match!) 

Note that the subject of the embedded clause being relativized is humaa 
'you' whereas the head NP subject is ?antumaa 'you'. The former is a 
third person pronoun whereas the latter is a second person pronoun. It 
would be interesting to see how the so-called copying-deletion analysis 
would account for this type of relativization. 


*I wish to thank Professor E. Bokamba for his generous support and 
invaluable advice throughout the writing of this paper. I am indebted 
to him for his time and consideration. I also wish to acknowledge Pro- 
fessor Jerry Morgan and Professor Yamuna Kachru for their helpful com- 
ments on earlier versions of this paper. All mistakes made are my own. 


'?' represents the glottal stop. 


9 represents the voiced pharyngeal. 


h' represents the voiceless pharyngeal, 


Arabic has one form of the definite article -al . However, there is a 

phonological assimilation rule which assimilates it with the following con- 

f+anteriofl „ ,, , 

sonants if they are , • Some of these consonants are: d, s, s, 
l+coronal — — — 

£.' £.' £.> ^""^ ^'-' '-"^* 


AISSEN, J. 1972. Where do relative clauses come from? In J. P. Kimball 
ed.: Syntax and Semantics, Vol. 1, pp. 187-98. New York: Academic 
Press . 

AL-HAMTHANY, A. 1967. Sarh Ibn Aqeil, ed. by A. Abdul Hamid . Cairo: Al- 
Maktabah Al-Tijar iyyah Al-Kubraa. 

ANSHEN, F. and P. Schreiber. 1968. A focus transformation of Modern Stand- 
ard Arabic. Language 44.792-97. 

AWWAD, M. 1973. Relat ivization and related matters in Modern Standard 

Arabic and Palestinian Arabic. Brown University Ph.D. dissertation. 

BAKIR, M. and J. Murtada. 1980. Aspects of clause structure in Arabic. 
Bloomington, IN: Indiana University Linguistics Club. 

BASHIR, S.M. 1982. A functional approach to Arabic relative clauses and 

implications for language teaching. Columbia University Ph.D. disser- 

BEESTON, A.F.L. 1968. Written Arabic: An approach to the basic structure. 
Cambridge: Cambridge University Press. 

BOKAMBA, E.G. 1976. Relativizat ion in Bantu languages revisited. In P. A. 
Reich, ed.: The Second LACUS Forum 1975, pp. 38-50. Columbia, SC: 
Hornbeam Press, Inc. 

BOKAMBA, E.G. and M. Drame. 1978. Where do relative clauses come from? 

In D. Farkas, W.M. Jacobsen, and K.W. Todrys, eds . : CLS 14, pp. 28-43. 

KEENAN, E. and B. Comrie. 1977. Noun phrase accessibility and universal 
grammar. Linguistic Inquiry 8:1.63-99. 

KILLEAN, C.G. 1972. Arabic relative clauses. In P.M. Peranteau, J.N. 

Levi, and G.C. Phares, eds.: The Chicago Which Hunt: Papers from the 
Relative Clause Festival, pp. 144-52. Chicago: Chicago Linguistic 

LEWKOWICZ, N.K. 1971. Topic-comment and relative clause in the Arabic 
language. Language 47.810-25. 

MORGAN, J.L. 1972. Some aspects of relative clauses in English and Alban- 
ian. In P.M. Peranteau, J.N. Levi, and G.C. Phares, eds.: The Chicago 
Which Hunt: Papers from the Relative Clause Festival, pp. 63-72. Chi- 
cago: Chicago Linguistic Society. 

RADFORD, A.T. 1982. A student's guide to transformational grammar. Cam- 
bridge: Cambridge University Press. 

SUAIEH, I.S. 1980. Aspects of Arabic relative clauses: A study of the 

structure of relative clauses in Modern Written Arabic. Indiana Uni- 
versity Ph.D. dissertation. 

studies in the Linguistic Sciences 
Volume 14, Number 2, Fall 1984 

Herbert Stahlke 

Phonological constraints on the lexicon of a lan- 
guage may describe a form which never occurs phonetically. How- 
ever, not all constraints on morpheme structure are lexical. 
Rule interaction in Ewe, affecting both vowels and tones, 
results in the neutralization of the phonological sequences 
-ee# and + psonorant"! [+h] I+h) #, the former with -ii#, 

ind + psonorant"! 
|_+voiced J 

and the latter with a [-H] [+H] sequence. Only nouns have a 
structure which participates in these neutralizations. As a 
result the Kpando dialect has no lexical entries for nouns 
ending in the two sequences that are neutralized. The 
elimination of these sequences by morpheme structure conditions 
rather than by rule interaction is costly and is contra- 
indicated by dialectal evidence. 


Phonological theory allows a variety of hypotheses con- 
cerning the nature of underlying phonological representations. 
One of these positions, the now-classic theory presented in 
Chomsky and Halle's The Sound Pattern of English , allows an 
underlying representation which exhibits properties of 
various superficial representations while remaining distinct 
from all surface forms. Such abstract underlying forms are 
argued for in Kenstowicz and Kisseberth (1977) on a number 
of grounds. One example of such abstract forms is cited from 
the American Indiana language Tonkawa, which has the surface 
root forms show in ( 1 ) . 

(1) A B C D gloss 

' hoe ' 












n tale 

■ lick' 

picn pcen picna pcena 'cut' 
(Kenstowicz and Kisseberth 1977:16) 

The underlying representations of the three root sets are 
(2) . 



' hoe ' 
' lick' 


' cut ' 

None of these underlying forms ever occurs as a surface form, 
but the surface forms cannot be derived without the informa- 
tion contained in (2) . In each case a vowel deletion analysis 
is simpler than a vowel insertion analysis since vowel inser- 
tion would require specifying the vowel for each syllable of 
each word, a condition not unlike writing rules to insert the 
vowels of English monosyllabic verbs. 

to c 
or t 

he CO 
es ma 
, for 
ed el 
r ly in 



n uni 

s on 



St of 

y als 


s true 



g mor 


e wit 
act u 
es of 
ider e 

o be 
pie , 
re (S 


h the 
1 and 
erne st 
the m 
d to h 
of Ewe 
ns in 
onant . 







, in 


r ea 










h ar 
al w 

ne ss 
s of 


e ne 
it a 
ot b 

a 1 

cif i 


a la 



ts f 

e ex 


c CO 

f oun 
or s 
al n 
er o 


age are 
ndi tion 
7) defi 
ge and 
cing re 
he cano 
d phone 
asal . 

ned wit 
ccur s a 

the claim 

s . These 
ne the 

dundancy , 
tical ly . 
g the 
I have 

hout an 
s a 

It has occasionally been observed that the general- 
izations captured by morpheme structure conditions are found 
also in the phonological rules. In vowel harmony systems, 
for example, restrictions found within morpheme boundries 
frequently extend to the vowels of derivational and in- 
flectional affixes as well. Fresco (1970) demonstrates that 
in some Yoruba dialects, such as Ifaki, the person and 
tense/aspect prefixes on a verb harmonize with the vowel of 
the verb, as in (3).^ 

(3) mo tl lo 
mo ti ri 

' I have gone . ' 
' I have seen . ' 


In other Yoruba dialects, including Standard Yoruba, vowel 
harmony is only a morpheme structure condition and does 
not spread to verbal prefixes. Thus 'I have gone' would 
be as in ( 4 ) . 


mo ti lo 

' I have gone . ' 

For the Ifaki dialect it could be argued that the morpheme 
structure condition functions as a so-called 'anywhere' rule, 
applying in an unordered fashion whenever its conditions 
are met . 

Kenstowicz and Kisseberth (1977) discuss such inter- 
actions between morpheme structure conditions and phono- 
logical rules as cases of functional unity or conspiracy. 
The grammar may contain otherwise unrelated sets of rules 
whose combined effect is to bring forms into conformity 
with a surface constraint. A case of this in Kpando Ewe 
is the Equal Height Condition on vowel sequences which 
requires that all word-final phonetic vowel sequences con- 
tain vowels of the same tongue height. This surface phonetic 
condition is the result of a number of independently moti- 
vated, unrelated rules that adjust the height of one vowel 
to another (Stahlke 1971b). 

in a 
to c 
in t 
to b 
def i 
ac ti 

c ture 
s lat 
r eff 
er tai 
his p 
ons a 
cal g 
s end 
s of 

e , a 

er i 
n ou 


nd c 
s on 
s . 

s of 
11 fo 
n the 

to p 

ain o 
and i 
s s of 
an be 
ly at 
Two e 
in a 

ns a 
, or 


n th 

d by 


be d 


re a 
, al 
ce p 



e , 

eract i 
oking . 
gy. A 
ns on 

al mor 
con of 

ed for 
t of o 
of sue 
ssed , 
and th 

on discuss 
That is , 

to apply 
ively are 
Iso, certa 
c sequence 
the gramma 
rules may 
1 forms fr 
may elimin 
pheme from 

certain p 
nted for b 

by morphe 
therwise u 
h rule-det 
one involv 
e other th 

ed so 
in ru 
s tha 
r . I 
om oc 
ate a 

y nor 
me St 
ing t 
e abs 



t CO 

t is 
ed s 
he a 

are , 

d by 
nf orm 



rule inter- 

y stematic 
bsence of 

of noun 


- sonorant 




The Absence of Cee 

In most works on Ewe, the third singular clitic 

pronoun is given as short e^. In the Kpando dialect the 

subject pronoun will always become short e^, and the object 

pronoun will do so after short a^ and short o_. 

In all Ewe dialects, the following forms are found. 

( 6 ) Underlying form Assimilated form 

(i) fi + e 'steal it' fii 

( ii ) wu + e 

( iii ) dze 

( i V ) po + e 

'kill it' 

' spoil it ' 

' beat it ' 

dzee or dzii 

poe or pui 

The general tendency, then, is for the pronoun to assimilate 
to a high vowel in a verb stem. Thus in (6) we find sequences 
of high vowels and sequences of non-high non-covered vowels, 
but there are no vowel sequences in which we find both high 
and non-high vowels. The basic rule in (6) , then, seems to 
be something like (7) . 



phigh n 





That is, short e becomes short _i after a high vowel. The 
second vowel change attested in (6) is the raising of non- 
high, non-covered vowel sequences to high vowels, giving 
dzi + i and pu + i from dze + e and po + e respectively. 
We will formulate stem vowel raising as in (8) . 


Phigh 1 
trcover edj 



[ -covered] 


Rule (7) must apply only in word-final position. Otherwise 
all non-final sequences of non-covered vowels will be99me 
high. This is wrong since such sequences as ie. in asieke 
'nine' and mieyi 'we went' never become long ii. The final 

form of (7), then, is (9) 


Final vowel assimilation 
V --^ [+high] 

Pback 1 

/ V 




There is also a rule of e^-lowering which all dialects 
seem to have, by which short e^ become short e^ when following 
the low vowel a. This is coupled with a rule raising short 
a to short £ , so that no Ewe dialect for which we have in- 
formation has the word-final vowel sequences a^ or ae_ in 
surface forms. ae will be present in underlying forms as 
one of the sources for long £_£ in all dialects. e^-lowering 
operates across word boundaries; a^-raising applies just 
within a word. Therefore the two rules must be stated 
separately as (10) and (11). 


v> [+covered] 

y [-low] 

/ [+low] 


These two rules, then, account for the assimilations in (12) 
producing forms which are found in all Ewe dialects. 

(12) Underlying form 

(Ja + e ' cook it ' 
na + e ' give him ' 
ll + e 'cut it' 

Assimilated form 


The treatment of the third singular object pronoun in 
Kpando differs from other dialects only in that coveredness 
assimilation lowers short e^ to short e after short o_ as well 
as after short a. Thus where other dialects have (13i) , in 
which the underlying form is unchanged in the surface, Kpando 
has (1 3ii) . 

(13) (i) w5e 

' do it ' 
' s tab i t ' 
'send it' 
' find it ' 



'do it' 


' stab it 


' send i t 


'find it 


The surface vowel sequences which will be found in 
Kpando verb + third singular object pronoun constructions 
are given in (14) with their underlying sources and de- 
rivational paths. 

(14) ii,r- ie by (9) 

""■^e by (8) and (9) 

ee ee 

ee ae by coveredness assimilation and 

oe oe by coveredness assimilation 

oe oe 

uir oe by (8) and (9) 


^Nae by (9) 

If a noun ends in a final vowel sequence, the sequences 
which are found at the systematic phonetic level in Kpando 
are as f ol lows : 

(15) ii 


' digging stick ' 


' cotton ' 


' nut ' 


' root ' 


' head ' 

> / 


'pool, body of water 


' nes t ' 


' cloth' 


' gourd ' 


' mortar ' 



' mouth ' 


' seed ' 


' soap ' 


■ antelope ' 



' palm of hand ' 


button ' 

ui tui ' broom ' 

/ / 

kui 'native money' 

As (15) shows, Kpando has no nouns ending in long ee. 
The absence of this vowel sequence can be accounted for 
synchronically in either of two ways: either rule (8) 
raising short -e to short i^ before any [-covered] vowel is 
obligatory in nouns, even though it is optional in verb and 
object pronoun sequences; or long e_e become long ee by a 
number of rules for ^-lowering discussed in Stahlke (1971b). 
If the former analysis is adopted, we have a complete 
synchronic merger of the underlying sequences long ii, ie 
and long e^ in nouns. Rule (8) will raise the first~ 
segment of long e_e to i^, and then (9) will raise any final 
£ which immediately follows _i. Thus there would be no means 
of distinguishing the long i^ sequences which are underlying 
from those which arise through (8) and (9). If such an 
absolute meutralization as we have proposed actually occurs 
in Kpando, there is no longer any basis for claiming that 
Kpando has underlying long e_e or i^ sequences in nouns 
since these sequences will always become long ii. The 
proposal that long e_e becomes long e e seems obviously 
wrong since e^ does not lower in a vowel sequence unless 
the other vowel of the sequence is [+covered] that is e, 
£, or a. Long e_e, then, is blocked from lowering unless 
we posit a special rule, applicable only to nouns, to 
lower long e^ to long ^e^. Such a rule would be arbitrary 
since there is no evidence for it beyond the problem it is 
intended to solve. Further evidence that long ee is not 
derived from long ee^ is the fact that the two do not 
alternate. Short e does not lower if the adjacent syllable 
contains a high vowel, and so we would expect that, if nee 
and kee are underlying nee and kee , they will remain so if 
followed by v_i, the diminutive clitic, producing *nee + vi 
'small nut' and *kee + vi 'small root' , instead of the 
correct forms nii + vi and ke £ + vi . Thus not only does this 
second analysis require an arbitrary rule, but this rule 
will apply in an environment where the corresponding rule 
for short vowels cannot apply. 

The problems involved in the second analysis would 
suggest that underlying long ee has merged with long ii 
and that synchronically Kpando no longer has long ee aT 
an underlying vowel sequence, were it not for the Tact 
that in nouns Kpando long ee usually corresponds to long 
ee in other dialects (16). 














' calabash ' 
' root ' 
'witch ' 
' sand ' 


ee , 


nei t 













is d 





r nat 
r ly i 
t ion 


r lyi 


er iv 
e ' s 
s su 
do h 

ica 1 1 
ion b 
of ou 
ion , 
ng lo 
is w 
ower i 
it wo 
ng se 
ugh a 
ves u 
his to 
aly se 
ed el 
s of 
(19 ) 
as on 

(17) (i) 

y , then , 
is no lo 
etween 1 
r synchr 
we are f 
ng ^^ in 
ated by 
here Ion 
ng since 
not lowe 
uld be m 
gment in 
11 other 
s with o 
r ical ly , 
d as ae , 
long ee 
to this 
ly the f 


Ewe Ion 
nger a s 
ong e_e a 
onic sol 
orced to 


a morphe 

g e_e_ com 

there a 

r elsewh 

ost unus 

a langu 

vowel s 

ne alter 

the res 

the onl 

in the g 

in Kpand 

agbe^ ' 

ana ly si 

o 1 lowing 

g ee 
nd lo 

me bo 
es f r 
re no 
ere i 
ual f 
y seq 
r amma 
o nou 
s . U 


has becom 
onic proc 
ng ZZ_ are 
s stands 
, where t 
om . I t c 

n the syn 
or a long 
n which 
both long 
e . After 
g vowel s 
uence fro 
r. Thus 
ns is the 

plate ' , 
nder lying 
1 sequenc 

e Kpa 
ess s 
f oun 
up un 
he tw 
The n 

ee t 


m whi 
ly, t 
es . 






o vo 


ic p 
o be 
£ is 

ce 1 
ch 1 
ly in 



e is no 


s ince 
honology , 


not , 
t forms . 

ong ee_ 
ong e_e_ 
g all 

ae . 
a + e , 

hen , 



Those in (17ii) are realized as in (18) 
(18) ii ui 

oe or ui 



Our analysis of long ee as the surface representation of 
ae enables us to accourTt for an otherwise inexplicable gap 
in the set of surface vowel sequences. Short a^ is the only 
covered or back vowel which does not appear before a non- 
low front vowel in the surface phonology, although we 
would expect the surface sequence ae_. However, rule (11), 
which raises a^ to £ before £, eliminates all surface 
instances of ae . 









it do 

and 1 

e^- low 

but i 

the s 



er a 
ns t 
as t 

es n 
er in 
t is 
, an 

e , as 
n unde 

ii . 
g vowe 
ake pi 
he und 
ot hav 
£ E are 
g rule 

r e s tr 
nee oe^ 
s . Pe 
d not 

we me 
in Kp 
ace , 
er ly i 
all a 
e unc 
of c 
ic ted 
as o 
ki , t 
el sew 

g lo 


so t 
ng V 

f ou 

hen , 

, we 
ng e_ 
n so 
he s 
owe 1 



e wou 

s app 

me di 

ur f ac 

s . P 


ed e^- 


ss as 

s , si 


e- lo 

, th 
Id r 
e vo 
eki , 

+ o 
wer i 

e di 
ts , 
wel s 

h Kp 
s ha 
b j ec 
ng o 

n abl 
n Ion 
the o 

ve th 
on as 
e (19 
t pro 
nly a 

ive -e^ IS 
e to test 
g ee or be- 
all seven 

ass imi- 

be the 
ther hand , 
has , but 

so long ee 
e same 

Kpando has , 

) lists 
noun con- 
fter the low 

and s 
As we 
a sin 
as is 
in no 
r aisi 
the 1 
the i 
and ( 
by ph 
be re 

The diac 
s an int 

sible to 
gle morp 
in all c 
nat ions . 
impl ies 

the eas 
same rul 
uns . To 
fore , wo 
ng. The 
exicon c 
tion, bu 
nf ormati 
9) . Sin 
ce of lo 
ture . T 
dundant . 

t rea 
ed ab 


e in 
e abs 

uld c 

t sue 
on wh 
ee (8 
ng ee_ 
al ru 
he mo 

c lower 
nalysi s 

ing pro 
ove , th 
and und 
be rais 

fact t 
there w 
verb + 
that Kp 
nee of 
be capt 
h a con 
ieh mus 
) and ( 

from t 
lea rat 


e vo 

er ly 
ed t 
ob j e 

et t 
t al 
9) a 
he 1 
s tru 

of Kp 



of 1 
ing 1 
o Ion 
ct pr 

he im 

ee f 

by a 
on wo 
so be 
re ne 
etur e 

ee a 

g ii^ 

be a 
no 1 


s ta 
n is 
by c 


s ae 
al r 
ng r 
i i , 

wi t 
a vo 

n se 
, it 


g e_e to long 

in eonformi 
do phonology 
epresentat io 
ules make it 
ng long e_e i 
since the fo 
h no tell-ta 
wel-rai sing 
el to be rai 
quences , but 

ii and long 
ee in its le 
n of vowel- 
le morphemes 
e structure 

to duplieat 
in rules (8) 

seems that 
be accounted 
tions on mor 
on would sim 



sed , 





Long High Tone after Noun Stem Voice Obstruent 

In noun roots with long vowels, Ewe allows the surface 
tone patterns shown in (19) . 

(19) (i) H H 

* t 




' skin ' 

' palm nut ' 

' axe ' 

' sand ' 




M H 


' fish' 

f ii 

' digging 


ama a 

' greens ' 

~ t 


' bridge ' 

L H 


' plate ' 


' cloth ' 


' mud ' 


' horn ' 

A number of facts strike one immediately from (19) . 
First, all stem-final long vowels end on a high tone. 
Second, HH and MH contain no nouns with a voiced ob- 
struent as a stem consonant, and LH contains nouns only 
of that shape. This distribution is consistent with Ewe. 

Although there exists as yet no satisfactory expla- 
nation for the absence M and L in final position, there : 
obviously no such restriction on the first stem vowel. 
Only if that vowel is preceded by a voiced obstruent do 
we find a restriction on tone, since Ewe nouns do not 
generally allow a high tone after a voiced obstruent 
stem consonant. In compounds, howevejr, where the noun 
prefix has been deleted, a high tone will be found after 
some stem-initial voiced obstruents, as in (20) . 

(20) (i) nudaz^ 

'pot for cooking food' 

cf. eda nu ze£me 'He cooked food in a pot 

kosf 4a ' Sunday ' 

(kosi (day name) + daa 'date') 

alalahe 'knife for cutting meat' 

/ / — / 

cf. etse hee la ala 'He cut meat with a 

kni f e . ' 


(ii) gagoo' 'oil drum' 

(ga 'metal' + goo 'gourd') 

tsinodzoo 'horn for drinking water' 

cf. ets£ dzoo no tsi 'He drank water with 
a horn . ' 

The alternation between high tone and low-high rising tone 
requires a rule of tone- insert ion which, together with 
other rules, will operate on underlying stems of the form 
(21i) to produce the surface forms show in (21ii) . The 
prefix of any obstruent- ini t ial noun will be realized on 
a low tone. This lowering does not occur before a stem- 
initial sonorant. 

(2 1) (i)P f-H 

3+ 1-sonorant 

[+voiced J 

( i i ) Underlying form 

acja ' sugar cane 


Derived form 













A rule of low tone assimilation (LTD will then apply to 
nouns of the shape (22i), accounting for the surface forms 
show in ( 2 2ii ) . 


' witch ' 


' mat ■ 


' arm ' 


• forest ' 


' dog ■ 


'oil palm ' 


' throat ' 


' work ' 


' sky ' 


' pot ' 


• goat ' 

(22) (i) 

+ L 



- sonorant 
+ voiced 


( ii ) Underlying form 

adzoeN 'tsetse fly' 
avoo ' cloth ' 

Surfa ce form 

In order to account for these and a number of other al- 
ternations, the following rule of tone insertion (23) 
is needed. 


' saliva ' 


'garden ' 


' native land 


' horn ' 


' gourd ' 


---•? ;-H I / 


[ + L] 


+voiced 1 








bef o 



in t 


of t 



tion . 
s wi t 

ng mo 
re th 
d sho 
he le 
his s 
d app 

h an 

re t 
e fi 
r ten 

ear , 


te t 


1 is 





n is 



e 1 9 

nd v 

e th 
h to 






ne s 
y do 

e are 
ced o 
of su 
ed by 

e and 

thr e 
pref i 

not . 

n ra 

ch s 

h bl 

e vo 
e ab 
y ac 
wi th 


n th 


a ve 

rly in 


at t 


a sy 

d ins 

e Two 


e of 

ted f 

ns , i 

so fa 


ry 1 
g or 

a 1 
he s 

or o 
n wh 
il t 

le f 
a vo 
el C 
to t 
r de 
o un 
e ob 

es t ing 







ondi t ion 

wo by 


scribed . 



s truents 



a + CV a + CVV 

a+CV a+CVV Prefix lowering 

a+CVV a+CVVV Tone insertion 

a+CVV a+CVV Low tone 


by a mo 
vowel o 
stem CO 
would , 
LTI and 
be anot 

e lexical abs 
rpheme struct 
f a long vowe 
nsonant is a 
however, dupl 
the Two Vowe 
the grammar 
her case , ver 
es discussed 
or sequences 
gical rules , 
ons . 

ence of 
ure cond 
1 noun s 
voiced o 
icate th 
1 Condit 
y simila 
earlier , 
from the 
rather t 

such f 


tem as 

bs true 

e info 

ion an 

ar i ly . 

to t 

whe re 


han by 

orms c 

nt . T 
rmat io 
d woul 
hat of 

the a 
on i s 


igh j 
his r 
n car 
d the 
to be 
erne s 

be described 
the first 
ust if the 
es tr ict ion 
ried by 
refore com- 

appears to 
long e_e 
e of certain 

captured by 
t r uc ture 

of n 












or c 





The ph 
be posi 

ent seq 
ents . 
r lying 
ot be d 
h are i 
esent g 
cture c 


s di 
to u 
It a 

al f 
nts . 
n fa 
s . 

on , 

es , 


ct s 
in t 
t ion 

t mo 
on , 
h do 
rs t 
ke t 


ch c 
he f 


es n 
n is 

i s , 
g f o 


d in th 

are di 

ted Kip 

ly, tha 

ot occu 

that t 


rms rat 

while o 

in (2) 

y by an 


y by mo 


cally a 

of phe 


1 s pap 
ar sky 
t no u 
r phon 
hese c 
her th 
ne can 
, wher 
ogy of 
may we 
d by t 

er a 

to p 




an t 

e a 

a 1 

11 p 

he E 

re a 
m the 
ces t 
o und 
repr e 
we mo 


abso lute 

1 ve 

er lying 

sentation ; 

e rules 
t forms 
of rule 
d above 

(25) MSC (Positive) 

+ Cons 
<.N o u n^ 







(25) says that a root begins with a consonant and possibly 
a liquid or the palatal glide. It contains one vowel, 
and, if a noun, may have a second identical vowel with a 
high tone, and it may end with a nasal. It would be 
possible to add the two conditions in (26) to block the 
sequences under discussion. 

(26) (i) MSC (Negative) 

+ C 

- low 
■ covered 

- low 


(ii) MSC (If-then) 


i -son 
j + voiced 

T (C) 



However, these 
grammar, and d 
are neutralize 
rules. That i 
is a function 
It is not a fa 
the canonicity 
uliar rule int 
from occurring 
dialects of Ew 
would make Kpa 
permit lexical 
in which case 
Most dialects , 
(8) , and this 
expect a langu 
were to change 
in long ee wou 
these sequence 
appear to make 
would a rule c 
change would a 
broaden the no 
constraints on 
interactions a 
tain the obser 
more easily th 
the source of 

conditions are 
o nothing more t 
d out by indepen 
s , the absence o 
of the interacti 
ct about morphem 
s about morpheme 

of such shapes, 
eractions discus 
In fact, such 
e. A change in 
ndo more like ot 
ization of dimin 
nouns ending in 

for example, do 
is clearly an un 
age to modify di 

in the Kpando d 
Id be permitted, 
s in the present 

that change mor 
hange be require 
Iso be required, 
tion of canonica 

morpheme struct 
s well. Such a 
vation that phon 
an morphological 
morphological ch 

complex , 
han bloc 
dently m 
f the se 
on of ph 
e struct 

struc tu 

and i t 
sed abov 

forms d 
any of t 
her dial 
utive fo 
- ie or - 

not hav 
ialect , 

form of 
e comple 
d, but a 
Thus i 
1 shape 
ure, but 
change e 
etic cha 

ange . 

k se 
ot iv 
ure . 
re w 
is o 
e th 
o oc 
ee w 
e St 

X , s 

t is 
to i 




at p 


em V 


s to 

di a 


es u 



to th 
ces w 

in qu 
al ru 
1 oth 

the p 
in ot 
es , w 
Ewe , 
ng in 

t one 
f thi 
ms en 


de no 
s to 


f req 




les . 




t them 






r . 


s rule 






y to 

t just 

of rule 



Epi log 


(cf . , e 
of vowe 
on the 
not wit 
from a 
pre sent 

nee thi 
repr ese 
na has 
•g. , ci 
ed in t 
al in n 
ge woul 
1 seque 
other h 
hout it 
high to 
ed as i 

s pape 
ntat io 
been a 
ement s 
hi s pa 
ature . 
d ar i s 
nces . 
and , i 
s prob 
ne to 
n (27) 

r was written, the use of autoseg- 
n for tone and other non- segmental 
rticulated in considerable detail 

and Keyser 198 ) . Both phenomena 
per could be represented as auto- 
It is not clear, however, that any 
e from the autosegmental treatment 

The autosegmental treatment of tone, 
s more immediately attractive, but 
lems. Autosegmentally , the change 
a rising tone in (24) coulb e re- 

(27) (i) V C V 

V C V V 

a d e — -' 

a d e 

/ I 
/ ! 

(ii) V C V V 

a d e' 

1 !\ 

L M H 

V C V V 

^ a d e 

M H 

The autosegmental representation seems to eliminate the 
problem of tone insertion, treating it rather as the more 
natural tone spreading (Human and Schuh 19 ). However, 
the stem vowels of the derived forms in (27) are both 
long and are furthermore of the same length. Thus rule 
(27i) must also insert a V element on the structural, 
CV tier. It is not clear that such an autosegmental 
analysis provides any greater simplicity than that pro- 
posal earlier, although the structural, rather than 
phonetic, treatment of vowel length seems more elegant. 
The auto segmental analysis will, however, provide a simpler 
and more elegant treatment of derivations like (24), where 
Low Tone Insertion produces a three-vowel sequence that must 
then be shortened to two under the Two Vowel Condition. Tone 
spreading does not generate the extra vowel, and the Two 
Vowel Condition becomes a function of canonical syllable 
shape captured on the CV tier. 


Both Ewe and Yoruba exhibit three surface tones, 
indicated by an acute accent for high tone, a grave 
accent for low tone, and a macronor diacritic for a mid 
tone . 

In the discussion following the presentation of 
this paper, Dr. Hounkpati Capo suggested that the under- 
lying form of the third singular object pronoun be i^, not 
e^. The strength of his proposal lies in the simplifica- 
tion it allows in phonological rules. Specifically, rule 
(8) , stem vowel raising could be restated as (i) 



be e 



ly in 








to e^ 





g vo 
ob j e 
con , 
be n 
h ex 
, as 

1 -h 
— c o v 


(9) , 
es . 
wel s 

ct is 

, not 

ing t 

ent a 




, si 
r , t 

e on 


t to 
he f 
s of 


[+high] ## 

al v 



t , i 




ly o 


to a 






the f 
t for 
ne en 
) , or 
ion o 
and o 







, th 
t fo 
d ha 

f (1 
e ( 
s th 

mi lat 
us to 
sub j 
1 dif 
is re 
e rev 
r the 
y mot 
ve to 
e se p 

ion , 
1 wo 

ec t 
er se 



) in 

uld a 
it di 
(e) a 
nee b 
ed f o 

of r 
ms in 
ed ru 
revi s 
e to 
d , Kp 


d appa 
f f eren 
nd ob j 
e tween 

to a 
r thir 
ule (9 

le (10 
ed to 
ando e 

nd i s 

rent ly 

di s- 

t under- 
ect (1) 

sub j eet 

) would 
and (13ii) 
) , 

lower i^ 
£ to e_, 
xhibi ts 
tion . 


ANSRE, Gilbert. 1961. The Tonal Structures of Ewe. 

Hartford Studies in Linguistics, No. 1. Hartford, 

Conn: Hartford Seminary Foundation. 
CHOMSKY, Noam, and Morris Halle. 1968. The Sound Pattern 

of English. New York: Harper and Row. 
CLEMENTS, Nick, and Samuel Kayser. 1983. CV Phonology: A 

Generative Theory. Cambridge: The MIT Press. 
FRESCO, Edward. 1970. Topics in Yoruba Dialect Phonology. 

Supplement to Studies in African Linguistics, Volume 

1. Los Angeles: UCLA African Studies Center. 





ICZ , 

, Her 
ds. ) , 
ngui s 







sser t 
, Ric 

M. , 
in K 

es i 
h In 
e 43 

es : 
ael , 

wa ' . 
er s 
n Ew 
ve P 
n . 


Russell G. Schuh. 1972. Universals 
Evidence from West Africa. 
iry 5:81-115. 
and Charles Kisseberth. 1977. Topics 
Theory. New York: Academic Press. 
971a. 'On the Status of Nasalized 

In Chin-Wu Kim and Herbert Stahlke 
in African Linguistics. Urbana: 
arch Incorporated. 

Surface Conditions on Vowel 
e'. In Charles Kisseberth (ed.). Studies 
honology. Edmonton: Linguistics 
orated . 

Topics in Ewe Phonology. UCAL doctoral 

967. 'Redundancy Rules in Phonology' . 

Studies in the Linguisti c Sciences 
Volume 14, Number 2, Fall 1984 

Aleksandra Steinbergs 

This paper describes some of the borrowing processes which take 
place in Tshiluba, a Bantu language of Zaire, and uses the data to 
explore contemporary generative phonological theories of borrowing. 
The data indicate that modifications in the shape of loanwords are 
not solely the result of the application of synchronic phonological 
rules of the borrowing language, but may derive either from morpheme 
structure constraints or as the result of universally common phonological 

Words which are borrowed into a language can be incorporated into 
the system of that language to differing degrees. This paper will briefly 
mention some of the factors that contribute to these differences of degree 
in loanwords in Tshiluba (a Bantu language of Zaire, principally spoken 
in the province of Kasai), and will go on to describe some of the processes 
which change the phonological shape of a loanword when it is borrowed into 
this language. 

1 . Degrees of incorporation 

In Tshiluba the differences in the degree of incorporation depend 
on factors like the age of the loanword and the amount of education of 
the speaker. In general speakers with little or no knowledge of foreign 
languages will use fewer loanwords of the type which deviate significantly 
from the canonical form of native words; likewise, older loanwords will 
normally show a higher similarity to the structure of native words. 


An examination of the morphological structure of Tshiluba reveals 
a pattern which is characteristic of Bantu languages: the system of noun 
class prefixes. Tshiluba assigns all nouns to one (or more) of eight classes. 
These noun classes are lexically determined, although in a few of them 
traces of semantic categorization still remain. The marker of each noun 
class is a pair of prefixes, one denoting singular, the other plural. 
A few examples of the native Tshiluba nouns, which illustrate this system, 
are shown in ( 1 ) : 

(1) class 1 mu-/ba-2 class ^ N-/N- 

[mukazl] woman [nzorlo] chicken 

[bak^l] women [nzc):l3] chickens 

[mulunda] friend [mpliku] rat 

[balundk] friends [mpbkil] rats 


class 7 <5:^-/bi- 
[cimuma] fruit 
[bimuma] fruits 

class 8 ka-/tu- 
[kambisi] cat 
[tumbisi] cats 




little woman 
little women 

Note that the prefixes of class 8 may also act as diminutive markers; -kai\ 
'woman' can, thus, appear in class 8 as well as class 1. 

In Tshiluba, one sign of a loanword which has been nativized to some 
extent is the presence of a noun class prefix. Another marker of (relatively 
strong) nativization is the avoidance of consonant clusters: native Tshiluba 
words have X but no r. consonant clusters except for sequences of nasal 
+ consonant. Both markers of nativization are seen in the loanwords in 

(2) [clsikiti] biscuit 

Neither of these markers is seen in (3); 

[bisikiti] biscuits 

(3) [krose] 

crochet (noun) 

As well as lacking these indications of nativization, the example in (3) 
also has a definite marker of foreignness: it contains an r. Native Tshiluba 
words have 1 but no c- In fact, the only real difference between the Tshiluba 
form of (3) and the word in French is the presence of tones. Moreover, 
even the tone patern (low tone on syllable not stressed in source language, 
high tone on stressed syllable) closely follows the intonational curve 
that the French word would have in isolation. It is clear that the example 
in (3) shows a very small degree of incorporation into Tshiluba. In fact, 
it is very possible that the pronunciation of this loanword in a Tshiluba 
sentence would be indistinguishable from the pronunication a Tshiluba speaker 
would give to an actual word of French inserted into a Tshiluba sentence 
(but, presumably inserted only if the hearers could also understand French). 

Thus, we see that the degree of incorporation of a set of loanwords 
may range from a Fremdwort barely distinguishable from the word in the 
source language to a Lehnwort so similar to a native word in phonological 
shape that it could be taken for native if the meaning so allowed. 

2. Phonological modifications 

2.1.1 My primary aim in this paper is to describe some of the phonological 
changes that occur to borrowed words and to describe the motivations behind 
such changes when they do, in fact, occur. Note that it will not be a 
counter-example to my claims if loanwords exist which do not undergo the 
processes that I shall discuss; these words I will merely consider as loans 


which have remained similar to the form of the source words, and, thus, 

not loanwords which are fully incorporated into the native phonological 

system of the borrowing language. As I discuss those cases in which 

modifications have been imposed on a loanword, I will also mention some 

of the generative phonological treatments of borrowing as they are applicable. 

Although Tshiluba loanwords I will discuss come from several different 
source languages, I have found no evidence that the borrowing mechanism 
of the languge treats words from different source languages in significantly 
different ways. For example, English [ey] and [e], French [e] and [e], 
and Portuguese [e] are all consistently rendered as Tshiluba [e]. Most 
of the example words come from French and English; these were borrowed 
in the late nineteenth century or in the twentieth century. Some of the 
loanwords are from KiKongo or Portuguese (often via KiKongo), and these 
seem to have been borrowed somewhat earlier, while a very few come from 

In native Tshiluba words one finds the consonant system shown in (4); 
sounds in parentheses are allophonic variants only: 






P "3 

In general Tshiluba loanwords retain the shape of the word in the source 
language as much as possible. Even strongly nativized words do not radically 
alter their shape in many cases. Thus, in all of the examples in (5) none 
of the source language consonants and few of the vowels are altered in 
the Tshiluba words: 






Tshiluba loan 













roll call 




pick (axe) 


Fr [llinEt] [luneta] glasses (sg.) 

Eng [lets:] [dylleita] letter 

There are really only two major kinds of modifications found in (5): the 
addition of a final vowel, and the addition of a noun class prefix. 3 All 
native Tshiluba words are vowel-final, thus, vowels have been added to 
sheet , Dick , glasses , and roll call . Normally, the vowel added in final 
position is a. or a copy of the vowel of the preceding syllable. The noun 
class marker prefixed to most loanwords is a nasal. This is a marker of 
noun class 3 (N-/N-), and the nasal prefix appears in the loanwords for 
sheet , pick , and roll call . However, if it is possible to analyze a noun 
as belonging to a particular noun class (that is, where the initial syllable 
is similar to some noun class prefix), it is usually incorporated into 
this class. Thus, [kaye] belongs to the i:a-/iu- class (this is class 8; 
note [tuye] 'notebooks'), and [lune'ta] belongs to the iii-/JJ- class (class 
>i) . The word for letter has been assigned to the second most common class 
for loanwords, the s^J.~/s&- class (class 6). 

2.1.2 In any description of the possible phonological shapes of loanwords 
one of the major difficulties is the following: what principle does a 
borrowing language follow in order to replace a source language sound which 
the borrowing language lacks? In other words, what happens when there 
is no sound which is immediately indentifiable as the same sound, or, at 
least, one which can be represented by the same phonetic symbol? The principle 
which has traditionally been appealed to is the one which Hyman (1970:8) 
calls the principle of "phonetic approximation". Basically, the principle 
proposes that a source language sound which does not occur in the borrowing 
language is replaced by the phoneticallv closest sound . The vagueness 
of this sort of principle does not much matter in cases where the substitution 
is obvious. For example, it is no surprise to any phonologist that Tshiluba 
loanwords may have an i where the source language word had an n. As we 
saw from the chart in (^), native Tshiluba words have no n; 1. is the only 
liquid. Thus, there are many examples like those in (6): 

(6) Fr [fu:r] [mfu:lu] (brick) oven 

rail (of train) 

Eng [brAs] [bulosa] brush 

However, another gap in the native Tshiluba consonant system is the 
absence of £. The principle of "phonetic approximation" does not predict 
whether [k] or [rj] will replace a source language ^, since both are equally 
close phonetically. In actual fact, the sound which usually replaces 4 
is [g], as seen in (7): 

(7) Eng [gowld] [qolo] gold 
Fr [gra:s] [gasa] grace 




















[gAvana- ] 






2.1.3 As well, this principle does not even begin to account for 
situations in which the forms of loanwords alter even though the borrowing 
language iiaa the source language sound. In an attempt to deal with this 
problem Hyman has suggested that "foreign sounds are perceived in terms 
of underlying forms" (Hyman 1970:19). In this view, the source word would 
be subject to modification by the synchronic phonological rules of the 
borrowing language." This approach does account for some additional 
situations. For example, in Tshiluba there is a widespread phonological 
process which palatalizes alveolar consonants before 1.*^ We find that 
some loanwords (which usually are also the ones which are strongly nativized 
in other respects) have been palatalized by this rule, as shown in (8): 


Eng - 






[ maslrjy 1 ] 










However, some loanwords do not undergo this rule. Often these are less 
fully incorporated forms. Alternatively, they may be forms used by more 
educated speakers, who adhere more closely to the shape of the word in 
the source language. Thus, we find unpalatalized forms, like those in 





body of Christians 


[mas in] 











2.2 In this paper I will describe in detail two cases of loanword 
incorporation in which the source words are modified even though the borrowing 
language has the source sounds in question. As well, both of these are 
cases where there is no synchronic rule in the borrowing language that 
could have effected these modifications. 

2.2.1 The first set of examples are shown in (10): 

(10) Fr [mustike :r] [muSete'kele] mosquito net 

(gaze k 

Ki Kongo 








[muse :te] 




box, bag 

All of these words contain a source language alveolar sibilant which appears 
as an alveopalatal sibilant in the Tshiluba loanword. These examples, 
unlike the ones in (8), cannot be handled by Hyman's approach (that is, 
by considering them as underlying representations to which synchronic rules 
apply); although there is a synchronic rule in Tshiluba which palatalizes 
alveolars before i, the examples in (10) occur before [e], and there is 
no rule which palatalizes sounds in this environment. There are, in fact, 
thousands of native words with unpalatalized sibilants before [e] and no 
words for which one would posit a rule of palatalization in this environment. 
Clearly there is no synchronic phonological rule that could have applied 
to palatalize the sibilants in (10). If we assume that alterations in 
the forms of loanwords are normally not random or haphazard, then clearly 
we must have some linguistically plausible basis from which to work in 
order to account for these forms." 

At this point I will digress to introduce the type of descriptive 
device I would utilize in this case; then I will return to these forms 
and posit the specific solution I have in mind. 

In a paper entitled "Loan words and abstract phonotactic constraints," 
Kaye and Nykiel (1979) present a refined version of Hyman's generative 
phonological borrowing theory in which they stress the importance of deep 
phonotactic constraints. I presume here that the terms that Kaye and Nykiel 
use, "deep phonotactic constraint" and "word structure constraint," are 
both essentially equivalent to the term "morpheme structure constraint". 
On the oter hand, surface structure constraints are strongly emphasized 
by linguists such as Shibatani (1973) and Picard and Nicol (1982). Certainly 
it is evident that phonotactic constraints (whether deep or surface) must 
play a large part in modifying the shape of some loanwords. For instance, 
in Tshiluba there is no evidence whatsoever for the existence of a synchronic 
phonological rule that inserts vowels in word-final position; nevertheless, 
it is very clear that conditions on the structure of words require that 
all words be vowel final." 

The phonotactic constraints that both Kaye and Nykiel (1979) and Picard 
and Nicol (1982) discuss are generally set up in the form of prohibitions, 
that is, as negative statements of what may not occur. As Kaye and Nykiel 
(1979:75) also point out, within one and the same language a particular 
negative constraint may be positively expressed in more than one way. 
Thus, in Japanese, the English source word strike has two expressions as 
shown in (11): 

(11) sutoraiki 'labour strike' sutoraiku 'baseball strike' 

Nevertheless, a language may prefer a relatively small subset of all of 
the linguistically possible positive expressions of a negative prohibition. 


Thus, in Tshiluba two major strategies are used to comply with the prohibition 
against consonant-final words: a final vowel is added which is a copy 
of the preceding syllable vowel, or the final vowel [a] (less commonly 
[i]) is added. Note that the vowels [u], [e ] , and [o ] are never final 
unless they are copies of a preceding vowel. 

Therefore, I suggest that one should go further than negative constraints 
to set up particular positive processes which would describe the actual 
observed modifications. These sorts of processes I have designated loanword 
incorporation processes. 

In the case of the data in (10), although there is no prohibition 
in Tshiluba against sequences of [se], nor any synchronic rule which palatalizes 
[s] before [c], this is certainly a universally common environment for 
palatalization. Although some of the loanword incorporation processes 
that I would posit for Tshiluba appear to derive from the synchronic system 
of the language, data such as those in (10) lead me to suspect that some 
modifications are also based on universally common tendencies such as 
assimilation. Thus, I posit that the examples in (10) are the result of 
a loanword incorporation process deriving (at least in part) from a universally 
common tendency toward palatalization, and that it is for this reason that 
the environment of the loanword incorporation process is extended beyond 
the environment of the synchronic palatalization rule. The synchronic 
rule for ^ would have the format shown in (12), while the loanword incorporation 
process would be formulated as in (13): 

V 1 







' 1 

The loanword incorporation process is, thus, more general that the 
synchronic rule (at least for £) , and is strongly supported on the grounds 
of phonetic plausibility and universal occurrence. 

2.2.2 The other instance I will discuss concerns the treatment of 
bilabial stops. Most commonly the bilabial stops in fully incorporated 
loanwords remain unchanged, as seen in (14): 









to baptize 








bolt, pin 


[pa tat] 






frying pan 





[apcl] [mpe:la] 

[pAinp] [mpompi] 


roll call 
(bicycle) pump 

However, in certain cases source language words containing a ^i have instead 
a voiceless bilabial fricative in the Tshiluba forms, as seen in (15): 



















Ki Kongo 




In native Tshiluba words [p] and [$] are allophonic variants, as shown 
in (16): 

(16) [lu$asu] 




red ant 

[ bi^mpE ] 

grinding stone 
well, good 

As is evident from these data, in native words [p] is found only after 
a nasal (which is always surface homorganic), while [$] is found in any 
other position possible for a consonant in the language. The simplest 
and most natural formulation of this allophonic relationship would be as 
a synchronic assimilation rule, as in (17): 


-> p / N 

Note, therefore, that there is no synchronic rule in the grammar which 
will change [p] to [$]. However, there does appear to be a phonotactic 
constraint against the voiceless stop in any but a post-nasal position. 

Hyman (1970) discusses several fairly precise principles that he uses 
to predict the shape of loanwords in Nupe. Although they are intended 
to be language specific and not universal, I find it interesting that the 
principle which applies directly to the above case predicts exactly the 
wrong result for Tshiluba. Hyman says that "when a foreign segment appears 
in an environment in which the equivalent native derived segment does not 
appear, then the form of the incoming foreign word is modified so that 
the structural description of that rule is met and the segment in question 


is then derived in the appropriate environment" (Hyman 1970:40). As we 
know, the appropriate environment for n is following m. If the Tshiluba 
data worked according to this principle, the form of the word would be 
modified (by the insertion of m) so that the sound in question (£) would 
now appear in the correct environment. In other words, Hyman' s principle 
would predict forms like those in (18): 

(18) '[dyikompo] cup "[nsumpu] 


Although Hyman' s principle does not sound implausible as a method for 
incorporating loanwords (and certainly does describe Nupe), it obviously 
does not work of Tshiluba. It would seem that much more specific descriptive 
statements are needed for the particular languages being described; loanword 
incorporation processes are precisely these sorts of statements. 

As we can see from the data in (14), loanwords in Tshiluba render 
a source-word £ as £ only if the word is also provided with a preceding 
nasal class prefix, as in [mpana] 'frying pan' or [mpe:la] 'roll call'. 
Thus, it appears that the representation of sounds in loanwords is dependent 
on their morphological classification into noun classes; that is, that 
loanword incorporations processes which are phonological are ordered after 
incorporation processes which are morphological. 

Since there is no synchronic rule that will change source language 
S} 3 into bilabial fricatives, and since there also appears to be a phonotactic 
constraint that prohibits a' s in non-post-nasal positions, I will posit 
a loanword incorporation processes that changes [p] to E*] in these disallowed 
cases. Remember that nasal + consonant clusters are the only consonant 
clusters permitted in native Tshiluba words; since syllables of the CqV 
structure (in all but quite foreign-sounding loanwords), I can set up the 
loanword incorporation process to occur when source language ji' s are 
syllable-initial in Tshiluba. In mc clusters, of course, the syllable 
boundary will always precede the m and never split the cluster. Thus, 
the process would have the form shown in (19): 

(19) p > $ / . 

3. Conclusion 

Both in the case of palatalization before [e] and the substitution 
of [$] for [p], it is clear that the synchronic phonological rules of the 
borrowing language are alone insufficient to offer a solution. I do not 
think that it is surprising that this is so, since each act of borrowing 
is a one-time only diachronic event, not an ongoing synchronic relationship. 
Certainly it is clear that conditions on the structure of words must play 
at least a partial role. I have suggested that the positive descriptive 
devices (which I designated loanword incorporation processes) are an appropriate 
means of describing the precise results of such conditions. As well, it 
appears that languages can have recourse to phonetically plausible and 
universally common processes (such as assimilation) for loanword modifications, 
resulting in loanword Incorporation processes which, I submit, are, therefore, 
quite natural. 



*This is an expanded version of the paper I delivered at the 1981 annual 
meeting of the Canadian Linguistic Association in Halifax, N.S. Some of 
the data derives from my work with two native speakers of Tshiluba, Remy 
Tshibangu and Benoit Tshiwala, for whose assistance I am sincerely grateful. 

^This is somewhat of a simplification, since even uneducated speakers will 
use recent loanwords which are not highly nativized. 

^High tones are marked ', low tones \ If no tone marks appear, I have 
been unable to discover the tone pattern. 

3of course, tones are added as well, but in general the stressed (i.e., 
most prominent) syllable is given high tone, the most prominent tone. 

"^Note that the tone pattern on prefixes of loanwords is sometimes altered. 
Normally all noun class prefixes have high tone. In loanwords, however, 
the tone pattern usually seems to be as follows: high tone on syllables 
stressed in the source language, otherwise low tone (except for some two 
syllable French-derived words which have all high tones). 

^Note that replacement of r is not obligatory, as seen in the word for 
•fork'. This, of course, is merely a less fully incorporated word. Also 
possible is a complete deletion of the c as in 'grace' and in [nsapitaj 
'chapter' from French r^hapltre [sapitr]. 

^However, this is not to say that ail of the phonological rules of the 
borrowing language must apply. The fewer the rules which apply, the less 
incorporated the loanword would be. 

■^In native Tshiluba the rule changes [t] to [S], [z] to [^] , [s] to [1&], 
[d] to [dy], [1] to [dV], and [n] to [ny] before [i]. Note, however, that 
only [dy] and [ny] are allophonic variants; /c z s/ must all be considered 
underlying since they can occur before vowels other than U J: l-Ca:yij 
•odd, uneven', [coiba] 'hit, strike', [d£cu] • ear^, [Senega] 'polish . 
[okasa:ma] 'leopard', [muSoiko] a kind of fish, [zorboka] 'be pliable , 
[zu:ka] 'get up, rise' . 

^Although only a handful of loanwords of the type shown in (10) occur, 
some explanation must still be offered. If the consonants had remained 
unpalatalized, they would be closer to their pronunciation in the source 
language. There can be no explanation in terms of the phonological rules 
of the language; there is no rule which palatalizes [s] to [S] before 
[e]; both [se] and [1&e ] are common sequences of sounds in native words. 

^There is no eivdence that this is clearly either a deep or a surface constraint 
in Tshiluba. Perhaps it is what Shibatani calls a M/SPC (a morpheme /surface 
phonetic constraint), that is, one which holds at both levels. 


I'hoDital . The final 
[1] becomes [d^] before [i] according to the rule described in footnote 


BURSSENS, A, 1946. Manuel de tshiluba. Anvers: De Sikkel. 

DE CLERQ, A. I960. Dictionnaire tshiluba-francais. Leopoldville: Imprimerie 

de la Societe Missionaire de St. Paul. 
HYMAN, L. 1970. The role of borrowing in the justification of phonological 

grammars. Studies in African Linguistics 1. 1-48. 
KAYE, J. and B. Nykiel. 1979. Loan words and abstract phonotactic 

constraints. The Canadian Journal of Linguistics 24. 71-93. 
MORRISON, W.M. 1939. Dictionary of the Tshiluba language. Luebo: J. Leighton 

Wilson Press. 
.1965. A textbook of the Tshiluba language. Luebo: J. Leighton 

Wilson Press. 
PICARD, M. and J. Nicol. 1982. Vers un modele concret de la phonologie 

des emprunts. The Canadian Journal of Linguistics 27. 156-169. 
SHIBATANI, M. 1973. The role of surface phonetic constraints in generative 

phonology. Language 49. 87-106. 
WILLEMS, E. 1955. Le tshiluba du Kasayi. Luluabourg: Mission de Scheut. 


Studies in the Linguistic Sciences 
Volume 14, Number 2, Fall iyS4 

Brent Vine 

From the time of the earliest descriptions of many East 
and Central African languages, one finds evidence for weakly 
articulated vowels referred to variously as 'semi-mute vowels', 
'vowel-colored breaths', or 'shadow vowels'. The precise phonetic 
nature of these vowels -- which may be simply voiceless, or which 
may involve various phonation types -- remains unclear, and de- 
serves further study. But apart from the intrinsic phonetic in- 
terest of these vowels, the available descriptions suggest phono- 
logical behavior that is also potentially of some importance. The 
Bongo-Bagirmi group of languages is particularly rich in phenomena 
involving shadow vowels, and these are described in some detail. 
Two patterns emerge: (i) in languages with both shadow vowels and 
normally articulated vowels on the surface (such as 'standard' 
Bagirmi), shadow vowels undergo processes (like deletion or assimi- 
lation) to which ordinary vowels are not susceptible; (ii) in lan- 
guages with no surface shadow vowels (such as Ngambay, and some 
dialects of Bagirmi), certain vowels -- which, predictably, cor- 
respond with shadow vowels in related languages -- behave as if 
they were shadow vowels (i.e. as in (i) above). Finally, evidence 
of this second pattern (which raises interesting questions about 
the linguistic significance of such vowels in underlying represen- 
tations) is presented from Alexandre's analysis of Bulu, in which 
apparently epenthetic 'buffer' vowels are in fact phonologically 
distinctive relics of the second vowel of original Bantu CVCV stems. 

0. According to Tucker and Bryan (1966:60), 'An outstanding characteristic 
of the BONGO-BAGIRMI languages is that the final vowel of certain words is 
semi-mute , i.e. pronounced so softly as to be hardly audible, and readily 
elided before following sounds'. This is a phenomenon which has been docu- 
mented since the earliest work on Bongo-Bagirmi, as can be seen in such 
descriptions as the following: 

(la) Delafosse 1897:19 (for Sara): 

Souvent en sara les voyelles finales sont excessivement breves; 
on pourrait meme ne pas les prononcer, et elles disparaissent 
parfois, dans les mots composes; comme ce n'est pas constant, 
et que beaucoup de voyelles finales conservent au contraire 
toute leur valeur, j'ai distingue les voyelles semi-muettes par 
le signe v-^ qui marque la brievete. Ainsi les mots debe homme , 
mon^ enfant, devront se prononcer presque deb , man, en faisant 
cL peine sentir la voyelle finale. Au pluriel on pourrait dire 
et ecrire: deb ge , mon ge. 


Compare Gaden 1909:8 (for Bagirmi) : 

(lb) Le signe / plac'S en avant d'une voyelle, indique que cette 

voyelle disparait des que le mot n'est pas isole ou en finale. 
Example : 

kag /a, arbre; kag muta , trois arbres. 
Quand ce signe est absent, le mot est invariable: 

kaga , panthere; kaga muta , trois pantheres. 

More recently^ cf. Caprile 1968:9 (for Mbay) : 

(Ic) En position faible ... la reduction des oppositions vocaliques 
est tres importante. Le relachement, la centralisation, la 
brievete et jusqu'^ un devoisement partiel caracterisent les 
realisations vocaliques dans cette position et rendent leur 
identification assez difficile ... 

But this phenomenon is not peculiar to the Bongo-Bagirmi languages: 
it has been observed in a variety of East and Central African languages, 
such as Teso, Burunge, and Galla, with documentation, again, from some of 
the earliest descriptions. For Galla, witness the following statement by 
Praetorius (1893:4): 

(2) Alle auslautenden unbetonten Vokale neigen zum Abfall. Sie konnen 
entweder 'ganz stummen sein, Oder doch so kurz gesprochen werden, 
dass sie nur noch ein Hauch zu sein scheinen, immerhin aber so, 
dass der Vokal noch kennbar ist'. 

The internal citation ('ganz stummen sein ... noch kennbar ist') indicates, 
in fact, Praetorius' debt to still earlier sources (in this case a previous 
description by Tutschek.) 

The cumulative impression left by informal descriptions such as these, 
as well as Tucker's more recent statement tnat the semi-mute vowels of 
Galla are 'voiceless and barely articulated' (1967:661), is that we have to 
do with phonetically voiceless vowels; compare also the traditional designa- 
tions 'semi-mute vowel' and 'vowel-colored breath'. But precise phonetic 
data on such Vs (including instrumental analyses) are so far lacking, and 
it seems preferable to adopt, at least provisionally, the phonetically 
neutral term 'shadow vowel' (due to Hilders and Lawrance 1956). Despite the 
lack of such phonetic data, however, I would like to suggest that the 
phonological behavior of these Vs, as seen in the available published de- 
scriptions of the languages in question, is potentially of great interest . 

The paper examines, then, the following areas: first, the behavior of 
shadow vowels (henceforth Vs) in synchronic phonological systems, particular- 
ly in Bagirmi; second, the behavior of Bongo-Bagirmi Vs from a historical/ 
comparative point of view; and third, a brief comparison of Bongo-Bagirmi 
Vs with similar data elsewhere in Africa. The aim of the paper is not so 
much to propose and defend 'unique' or 'correct' analyses of the data, but 



rather to characterize tne potential interest and importance of these 
neglected phenomena, if only as a stimulus for future work, both 
descriptive and theoretical. 

1. In Bagirmi, Vs are susceptible to certain phonological processes which 
do not apply to normally articulated Vs. First, Vs (the best examples in- 
volve /a/j assimilate to bilabial or palatal nasal consonants of following 
morphemes, as can be seen in the following alternations: 

(3) /aka/ 'see': m-aka '1 see, saw' i-aku-ma 'thev saw me' 
^ ^ X X - X • 

j-aki-na 'thev saw him' 


/ada/ 'give': m-ada '1 give, gave' n-adu-m ga 'he gave me' 

m-adi-n ga 'I gave him' 

/taoa/ 'work, do': ma m-taoa '1 do, did' taai-n 'do it I' 

X X 

Compare /aca/ 'cut', which does not have a V: 

(4) /aca/ 'cut': m-aca '1 cut' m-aca-na 'I cut it' (etc.) 

Before non-nasal consonants of following morphemes, and before th*^ 
dental nasal [n], Vs are deleted; compare the following forms from the 
paradigm of /ada/ ^give' with the assimilated forms n-adu-m ga 'he gave 
me' and m-adi-n" ga 'I gave him' in (3) above: 

(5) /ada/ 'give': n-aj-je 'he gave us' 

n-at-se 'he gave you (pi.)'" 

Some further examples are provided in (6) (note the morpheme-initial [n] 
in the first item; morpheme-initial [n] does not occur) : 

(6) mano 'head of cattle' — > man-ne 'cow' 

bisi dog' bls-ge (biz-ge) 'dogs' 

kaga 'tree' kag-ge trees' 

taoa 'work, do' taa-ki 'work I (^pl.)' 

nana 'tooth' nan-la 'chew' 

abe 'go, leave' ab-se 'carry oft' 

V-aeletion is in tact more general, and applies across word-boundary 
(including compound-boundary), even if tne following word begins witn a 
bilabial nasal : 


(7) kamo 'eye — » kam kaga 'leaf (lit. 'eye tree') 

toto 'hill' tot mbasa 'stone' (lit. 'hill small') 

X ' 

naba 'man' nab ga 'man who' 

taaa 'work, do' m-taa ga 'I have done' 
X " 

bogo 'steal' mala bog ja 'the master (who) stole the meat' 

kaga kag muta '3 trees' 

(There are no examples before the palatal nasal [n] , which does not occur 
word-initially. ) 

In contrast, ordinary Vs do not delete before following morpheme- or 
word-initial consonants (cf . also (4) above) : 

(8) kaga 'leopard' — > kaga-ge 'leopards' 
nanla 'chew' nanla-ki 'chewl (pi.)' 
ali (n. prop.) ali-ge 'people named Ali' 
mara 'crocodile' mara ngolo 'big crocodile' 
gwoto 'be absent' ne gwoto lolo 'he is not here' 

In Sara and Kenga, preconsonantal Vs behave somewhat differently: 
according to Tucker and Bryan (1966:63), the preconsonantal V deletes, as 
in Bagirmi, but this leads to a further development, namely the insertion 
of a syllabic nasal consonant, homorganic to the preceding consonant: 

(9) Sara: boba 'male' — ► babm-man 'bull' ('male-cow') 

Kenga: maka 'bell 

da-makn-kaba 'carnivore' 
( 'meat-belly-wild animal'' 

( 'meat-betly-wild animal') 
kos3 'food, eat' kosn kos 'to eat food' 

X I 

(Both Vs and Vs generally elide before following Vs.) 

A somewhat different situation is found in other dialects of Bagirmi, 
as can be seen from field work of J. Lukas, as reported in Tucker and Bryan. 
In the material designated 'Lukas 1' in Tucker and Bryan's corpus, from a 
single village, no Vs are recorded for many words which consistently have a 
V elsewhere in Bagirmi; these forms exhibit, instead, normally articulated 
Vs, as in the forms in (10) : 

(10) 'standard' Bagirmi Lukas 1 

kaga 'tree' kaga 

'ana 'sheep' ana 

imi 'locust' imi 

Iji 'urine' I j i 

muju muju 


oil ^ ' 

'AJnu 'nose' amu 


ulu 'swallow' ulu 

ele 'bird' ele 

dtbe 'person' debe 

k-abe 'journey' k-abe 

marjo 'head of cattle' mano 

oso 'bite' oso 

'abo 'hippo' afo 
X ^^ 

nono 'child' nono 

Nevertheless, such ' Lukas 1' forms undergo the same processes described 
for Vs above, while 'Lukas 1' forms which correspond to Vs elsewhere do 
not. For assimilation, consider the form 'bite' illustrated in [W): 

(11) Lukas 1 

ini 'snake' + oso 'bite' : 

m osi-na 'the snake bit him' 

For deletion, consider the forms in (12) : 

(12) Lukas 1 

/iniffoso+je/ — > m os-je 'the snake bit us' 

mano 'head of cattle' man-ne 'cow' 

debe 'person' deb-ge 'people' 

ulu 'swallow' m-ul ga 'I swallowed' 

In the dialect of still another village, designated 'Lukas 2', final Vs 
of any kind are altogether missing, as can be seen from forms like those in 
(13) (the corresponding forms from (10) above are repeated for convenience) : 


'standard' Bag 


Lukas 1 

Lukas 2 

kaga 'tree' 



'ana 'sheep' 
imi 'locust' 


\ / 



'/Mnu 'nose' 

V / 



ele 'bird' 



dtbe 'person' 



mano 'head of 




'abo 'hippo' 
nono 'child' 


> / 



Descriptively, then, we find in Bagirmi the following three synchronic 
situations : 

(i) In the dialect of what may be called 'standard' Bagirmi, Vs are 
phonetically distinct, on the surface, from other Vs. The fact that Vs are 
distributionally restricted, however, could be taken as an indication that 
such Vs are underlyingly not distinct from other Vs , and result from some 
low-level process (whether optional, obligatory, or 'variable') which applies 
to Vs in word-final position. One might claim, for example, that phonetic Vs 
arise automatically in unstressed word-final position, while final stressed 
Vs remain unaffected: thus Bag. ['kaga] 'tree' versus [ka'ga] 'leopard'. It 
appears, however, that in the Bongo-Bagirmi languages generally, 'stress is 
normally on the first syllable of disyllabic words, and invariably so when 
the second syllable contains a semi-mute vowel' (Tucker and Bryan 1966:64). 
Thus, while it is probably true that Vs are never stressed, most other final 
Vs are likewise generally unstressed, resulting in actual surface contrasts 
like Bag. ['kaga] 'tree' versus ['kaga] 'leopard'. Apart from stress (which 
seems, moreover, to have no linguistic significance whatsoever in Bongo- 
Bagirmi), there are no other known factors which might condition the realiza- 
tion of Vs as Vs. In similar fashion, one could claim that Vs are automatical- 
ly inserted , under certain conditions. There are two immediate problems for 
such an analysis: first, while the vowel quality of Vs is identical to that 
of the V of the preceding syllable in a large number of cases, such that one 
might wish to envisage (e.g.) a copying relationship, this is by no means a 
general fact; second, the tones appearing on Vs do not appear to be predic- 
tably related to any word-level features, including the nature of the pre- 
ceding consonant. But the major obstacle to this and similar analyses is the 
fact that Vs regularly undergo certain phonological processes which normal 
Vs do not. As an alternative, it may be necessary to consider Vs as under- 
lyingly distinct from other Vs, i.e. specified by some (perhaps laryngeal) 
feature or features, depending on their precise phonetic realization. This 
sort of situation, while typologically somewhat rare, is not excluded (given, 
e.g., the existence of systems with distinctive laryngealized or murmured 
Vs) ; one may further compare, in fact, the nasalized vowels of the Sara 
languages (a subgroup of the Bongo-Bagirmi languages), which must be taken 
to be underlying (i.e. not derived from sequences of oral vowel plus nasal 
consonant), a situation which I have discussed elsewhere (1978). 

(ii) In contrast with the 'standard' Bagirmi situation outlined in (i), 
the dialect referred to as 'Lukas 1' shows a more complex situation. Both 
Vs and certain normal Vs (the latter corresponding to Vs in 'standard' 
Bagirmi) undergo those processes restricted to Vs in 'standard' Bagirmi. It 
is difficult to know how to account for such a situation. On the face of it, 
one is tempted to claim that the state of affairs in 'Lukas 1' is an arti- 
fact of faulty description: perhaps the forms with apparent V for V (like 
those in (10) above) really have phonetic Vs , in which case 'Lukas 1' is 
identical with 'standard' Bagirmi. But there is evidence, as we shall see, 
in favor of taking the description of 'Lukas 1' at face value -- i.e., 
evidence that a functional equivalence, as it were, of Vs and Vs is possible, 
and linguistically significant. 



(iii) Finally, Vs are altogether absent in the dialect referred to as 
'Lukas 2', posing no particular problems for the synchronic analysis, but 
indicating certain interesting consequences for the comparative and dia- 
chronic analysis of Vs in the Bongo-Bagirmi languages, to which we now turn. 

2. From a comparative perspective, the Bongo-Bagirmi languages as a whole 
pattern, with respect to Vs, much like the dialect situation internal to 
Bagirmi, as described above. Thus, some languages contain Vs which behave 
much like those in 'standard' Bagirmi (e.g. Bongo, Baka, Yulu, some dialects 
of Sara, etc.; cf. n. 3), while others appear to contain no reflexes of Vs 
at all, as in 'Lukas 2' (e.g. most of the dialects of 'Sara proper', i.e. 
Tucker and Brvan's 'Sara Mbai'). But most interesting are those languages 
comparable to 'Lukas 1', in which Vs are phonetically nonexistent^ but in 
whirh certain phonological effects must be interpreted as traces of Vs. 

In Ngambay (cf. Vandame 1963, Thayer and Thayer 19711, no Vs are re- 
corded. Nonetheless, certain final Vs behave like Vs , from the standpoint 
of e.g. Bagirmi. Consider first certain facts involving word-final /-3/ 
in Ngambay. 

According to Vandame (1963:13), it is sometimes difficult to dis- 
tinguish a real word-final phonemic /-3/ from an automatic stop-release, 
particularly when the [-9] bears the same tone as the V of the preceding 
syllable. He proposes to make the distinction on the basis of the behavior 
of [-9] before following Vs, as follows: prevocalic V-elision is not ob- 
ligatory, especially in deliberate speech, and yet some [-9]s consistently 
elide before Vs , while others consistently do not. Those words in which 
[-9] always remains on the surface are taken to contain / -d/ , while the 
[-3] which consistently elides in other words is taken to represent a stop- 
release. Thus, the words for 'grain' and 'tree' show identical preconsonan- 
tal realizations, with respect to word-final [-3]: 

(14) [kanda to cTa] 'The grain is where?' 
[kagi to aa] 'The tree is where?' 

But their prevocalic realizations differ, as follows: 

(15) [kanda oso nan] 'The grain fell to the ground.' 
[kag oso nan] 'The tree fell to the ground.' 

Thus 'grain' is analyzed as /kand?/, while 'tree' is analyzed as /kag/, with 
a stop-release appearing in preconsonantal position, as in (14). 

But this analysis is somewhat misleading, once comparative evidence is 
considered. Vandame does not seem to have noticed that the word for 'tree', 
for example, corresponds to forms with V elsewhere in Bongo-Bagirmi, thus 
Bagirmi kaga . Bongo kaga , Yulu kage .^ We are dealing, then, in all likeli- 
hood, not with a simple stop-release, but rather with a phonological proper- 
ty of certain historical Vs, which surfaces in some environments. Vandame's 
analysis, if synchronicalTy correct, would thus be the result of a restruc- 
turing, whereby forms like *kag3, with -V, together with a prevocalic V-dele- 


tion rule, were reanalyzed as /kag/ , with a preconsonantal 3^-insertion rule. 
But it is at least worth considering the more abstract analysis, whereby 
surface [-9#] < *-;» and surface [-a#] < *-i are still underlyingly distinct, 
comparable to the situation in the ' Lukas T' dialect of Bagirmi, as outlined 

A second indication of the linguistic significance of Vs in languages 
which do not distinguish them phonetically concerns certain word-final tone- 
bearing sonorants in Ngambay. According to Vandame (1963:37-8), liquids and 
nasals in initial and intervocalic position automatically bear the tone of 
the following V. In word-final position, however, certain liquids and nasals 
bear unexpected tones. Vandame attempts to account for these by invoking 
sentence-sandhi (i.e. influence, in this case, from following V-initial 
words) or some sort of conditioned influence from preceding Vs. Unfortunate- 
ly, Vandame does not provide sufficient data to justify a detailed analvsis 
at this time. There is reason to believe, however, that the explanation for 
such 'unexperteH' tnnes -- once again, from a historical point of view -- 
is that they reflect traces of Vs. Thus Ngambay /man/ 'water', realized 
either as [man] or [man], is not to be separated from Bagirmi mane. Bongo 
min i , Yulu mim i , etc. Essentially the same process is quite transparent in 
the Ngambay first person suffixed pronoun, the only case, significantly, in 
which a tone-bearing sonorant is distinctive: thus Ngambay -m 'me' corres- 
ponds to Bagirmi -ma. (Note also that in this way, it is possible to make 
some sense of Tucker and Bryan's remark (1966:64) that in those dialects of 
Sara with phonetic Vs, no Vs are found after liquids and nasals.) 

3. Turning now to languages outside of the Bongo-Bagirmi group: there are 
indications that Vs represent an interesting 'areal' feature of sorts, which 
appears in several diverse forms. 

Very little material is available concerning Vs described as Vs through- 
out the rest of the linguistic area covered by Tucker and Brvan, namely the 
Vs asserted to exist in Burunge, Geleba, Krongo, tne Teda-Tubu dialects, 
bhilluk, Teso and Galla. The available evidence, however, indicates that Vs 
in these languages have essentially the same properties as Vs in the Bongo- 
Bagirmi languaees, namely (i) they are restricted to word-final position; 
(ii) there are no restrictions as to whicn consonants may precede them; (iii) 
they may be tonally distinct from preceding Vb in the word, although they 
are often identical in quality to the V of the preceding syllable; (iv) they 
tend to elide more readily than normal Vs; (v) they appear, in related lan- 
guages or dialects, either as zero or as normal Vs." Although Vs are evident- 
ly not phonologically distinct in certain of these languages, minimal con- 
trasts like those in (16) indicate that they may somehow be distinct at 
least in Galla: 

(16) 'person' ace. nama dat . nama (m.) 

'girl' ace. mtala dat. intala (f.) 

'this/these' ace. kana gen. kana (m.) 


But more interesting, from a broader, typological perspective, are 
languages with no trace of phonetic Vs , but which present phonological be- 
havior reminiscent of Vs , and which should perhaps be further investigated 
in this light. I would like to suggest that such a case is represented by 
Bulu, a Bantu language spoken in southern Cameroun, and thus roughly ad- 
jacent to the southernmost extensions of Bongo-Bagirmi territory. The facts 
in question have been discussed by Alexandre (1962) , on which the following 
summary presentation is based. 

Several groups of Bantu languages of southern Cameroun are charac- 
terized by a relatively high frequency of monosyllabic stems ending in a 
closed syllable, corresponding to CVCV stems elsewhere. Typical languages 
of this sort are Mbtnfc, Makaa, and Bulu. But, while e.g. Nlbfent and Makaa 
present, as a result, a high frequency of consonant clusters in actual 
speech, this is not the case in Bulu; consider the following descriptions 
(references in Alexandre 1962) : 

(17) 'A word ending in a consonant, when there follows another word 
beginning with a consonant, requires a slight vowel sound after 
it, making a little buffer syllable between the two words.' (Bates) 

'When a word ends in a consonant, closely followed by another word 
beginning with a consonant, Bulu speech very commonly inserts a 
slight vowel sound between.' (Good) 

Alexandre, however, has shown that this apparently anaptyctic 'buffer' V is 
not the result of a low-level epenthesis, but a reflection of a segment which 
is phonologically distinctive. IVhile the quality of these 'buffer' Vs in Bulu 
is nearly always identical to, or systematically related to, the quality of 
the preceding V, their tones bear no such relationship. Moreover, the tonal 
patterns of these Vs are functionally identical to those of the second V in 
CVjCV2 stems. Finally, Alexandre suggests that the Bulu situation may repre- 
sent an intermediate stage in a gradual attrition process affecting final Vs, 
with the end stage, namely zero, represented in languages like Mbfcnt or Makaa. 

This sort of situation is strongly reminiscent of the patterns outlined 
for Vs above, both with respect to their surface properties and distribution, 
and with respect to their phonological behavior, as exemplified in particular 
by the analysis of the so-called 'epenthetic' [-3"] of Ngambay, involving 
three-way contrasts like Bagirmi kaga , Ngambay (preconsonantal) [kags] , Sara 
kag 'tree' . 


This is a slightly revised version of a paper presented at the 10th 
Annual Conference on African Linguistics, held at the University of Illinois 
at Urbana-Champaign in April of 1979. An expanded version appeared in 1981 
(see the REFERENCES), to which the reader is referred for extensive discussion 
of some of the theoretical issues raised bv the data. 


In addition to the data in Tucker and Bryan 1966:59ff., material from 
the Bongo-Bagirmi languages has been drawn from the following sources: Gaden 
1909 (Bagirmi); Hallaire and Robinne 1959 (Sara); Caprile 1968 (Mbay) ; 
Vandame 1963, Thayer and Thayer 1971 (Ngambay) ; Santandrea 1963 (Bongo, 
Baka, Yulu, Kara). Tone is marked where known. For a more complete descrip- 
tion of these and other relevant sources, see Vine 1978:119-20. 

The regressive assimilations exhibited by such forms are optional, or 

rather 'variable', and depend on such factors as speech style, speed of 

utterance, etc. For a similar situation, see Bolozky 1980:797 with respect to 

regressive voicing assimilation in Hebrew. 

The same situation is found in several other Bongo-Bagirmi languages, 
including Bongo, Baka, Yulu, Kenga. 

Tucker and Bryan (loc. cit.) note that 'occasionally stress is heard 

on the second syllable, when the first one contains a central vowel', while 
citing as^an example Stevenson's description of Bag. kela 'snake' as phonet- 
ically [ks'la]. Such forms would constitute, then, the only class of excep- 
tions to the claim that stress in disyllabic words is initial. But one can 
also question the status of this class as 'exceptional', since the over- 
whelming majority of such C9CV forms (extremely frequent, for example, in 
Ngambay) have high tone on the final V. One wonders, then, whether the schwa 
of the first syllable (which characteristically has non-high pitch) and the 
concomitant high tone of the second syllable combine to give an impression 
of stress, which need not actually be present. It should be added that the 
distribution of /9/ (as seen particularly in Ngambay) is by no means restric- 
ted to this position (CaCV) ; therefore, it would be problematic to claim, 
without further justification, that this /3/ actually constitutes a 're- 
duction' of some full V, conditioned, in fact, by word-final stress. The 
problems associated with the status of I'd! in these languages, however, are 
complex, and cannot be considered further here. 

It is along these lines that certain Ngambay forms in Thayer and 
Thayer's material, with tone marked on word-final obstruents, are probably 
to be understood, e.g. kag 'tree'. A number of such forms are consistently 
marked in this way by Thayer and Thayer, despite their claim that only Vs 
and nasals bear tone (1971:3.24). Thus Bag. mosu , Sara (dial.) mase , but 
Thayer and Thayer m£s 'blood' for Ngambay; Bag. kese, but Thayer and Thayer 
k6s 'cough' , etc. 

With respect to the last feature: an interesting situation worthy of 
further investigation concerns the Vs in certain Nilotic languages. In Shil- 
luk, most verbs have the shape CVCV, whereas in (e.g.) Acoli, the correspond- 
ing verbs are of the shape CVCV, with a normally articulated final V. Never- 
theless, this final V, which has no lexical value, seems also to be phono- 
logically 'irrelevant': it never appears on the surface if a form undergoes 
any phonological process whatsoever, and may not belong in underlying rep- 
resentations at all. For details, see Crazzolara 1955. 



As a postscript, let me report the existence of independent work on 
issues very closely related to those treated above. In a letter dated July 
14, 1980, Gerrit Dimmendaal informs me that his paper 'Non-voiced vowels 
in Turkana: a Nilo-Saharan feature?' (in Collected Seminar Papers 1979/80 , 
ed. Franz Rottland: Dept . of Linguistics and African Languages, Univ. of 
Nairobi] argues that voiceless Vs in Turkana have phonemic status, on the 
basis of data concerning such factors as stress and tonal patterning. He 
notes, in addition, that his previous work on certain Nilotic languages 
(among them Teso and Shilluk, cf. sections 0. and 3. above, with n. 6) 
has led to similar conclusions. On the question of the phonemic status of 
voiceless Vs in a number of languages, see also my earlier discussion 


ALEXANDRE, Paul. 1962. Sur la voyelle suffixielle du bulu. Journal of 

African Languages 1.243-252. 
BOLOZKY, Shmuel. 1980. On the monophonematic interpretation of Modern 

Hebrew affricates. Linguistic Inquiry 11.793-799. 
CAPRILE, J. P. 1968. Essai de phonologie d'un parler mbay. Bulletin de 

la Societe des Etudes Linguistiques Africaines, no. 8. 
CRAZZOLARA, J. P. 1955. A study of the Acooli language: grammar and 

vocabulary. London. 
DELAFOSSE, M. 1897. Essai sur le peuple et la langue sara. Paris. 
GADEN, H. 1909. Essai de grammaire de la langue baguirmienne. Paris. 
HALLAIRE, J. H. and J. Robinne. 1959. Dictionnaire sara-francais. Koumra- 

Fourviere: Mission Tchad (Lyon). 
HILDERS, J. H. and J. C. D. Lawrance. 1956. An introduction to the Ateso 

language. Kampala. 
PRAETORIUS, Franz. 1893. Zur Grammatik der Gallasprache. Berlin: Wolf 

Peiser Verlag. 
SANTANDREA, S. 1963. A concise grammar outline of the Bongo language / 

A small comparative vocabulary of Bongo, Baka, Yulu, Kara. Rome. 
THAYER, Linda J. and James E. Thayer. 1971. Fifty lessons in Sara-Ngambay. 

Bloomington, Indiana: Indiana University Linguistics Club. 
TUCKER, A. N. 1967. Fringe Cushitic: an experiment in typological com- 
parison. Bulletin of the School of Oriental and African Studies 

TUCKER, A. N. and M. A. Bryan. 1966. Linguistic analyses: the non-Bantu 

languages of North-Eastern Africa. Oxford: Oxford University Press. 
VANDAME, Charles. 1963. Le ngambay-moundou : phonologie, grammaire et 

textes. Dakar: Institut Francais d'Afrique Noire, memoiro no._69. 
VINE, Brent. 1978. Nasalization in the Sara languages. Afrika und Ubersee 

VINE, Brent. 1981. Remarks on African 'shadow vowels'. Harvard Studies in 

Phonology 2.383-427. 


Studies In the Linguistic Sciences 
Volume 14, Number 2, Fall 1984 

Jennifer J. Yanco 

The aim of this paper Is to show that In Spoken Lingala 
modifiers which have been categorized as adjectives can be 
more adequately subsumed under the category noun. Hagege (1974), 
in an investigation of adjectivals in some African languages, 
claims that there is no universal word-class adjective, and 
shows that in some languages words categorized as adjectives 
behave like verbs, while in other languages, words categorized 
as adjectives behave like nouns. He points to Bantu languages 
as examples of the latter type. The question of whether modifiers 
in Bantu languages should be considered nouns continues to be a 
problem. Although there has been much discussion of this question 
(see Welmers 1973: Chapters 9 and 10 for an overview), the issue 
remains unresolved. In this paper evidence from Spoken Lingala 
is presented — evidence which suggests that the words in question 
can be most accurately categorized as nouns. In Spoken Lingala 
any constraint on the distribution of modifiers is motivated by 
interpretation conditions on the constructions in which they 
occur and does not justify establishing a separate grammatical 
category adjective. 

The language investigated in this paper is the dialect of Lingala 
spoken in the urban center of Kinshasa, Zaire, and referred to here as 
Spoken Lingala. For increasingly large numbers of people living in 
Kinshasa, as well as in Brazzaville (Congo), Spoken Lingala is a native 
language. It is also widely used as a lingua franca among peoples living 
along the Zaire River. 

The origins of Lingala are unknown. Some attribute it to a now- 
extinct ethnic group; however, attempts by Guthrie to support this claim 
were unsuccessful (as reported by Bryan, 1959). Others see Lingala as a 
trade language which developed along the Zaire River as a result of contact 
among speakers of Bantu languages from Guthrie's Zone C, particularly 
Bobangi. Whatever its origins, Lingala has been an important contact 
language along the Zaire River since before colonial times. This 
importance continued during the colonial era, the turning point for Lingala 
coming in 1929, when it was declared the official language of the armed 
forces of the Belgian Congo. This had the effect of spreading Lingala' s 
sphere of influence throughout the nation. On two occasions previous to 
this (1918 and 1920), attempts were made to institute Lingala as the 
official language of the Belgian Congo (Sesep 1978:64). In more recent 
years, its importance has increased greatly as it has become the native 
language of expanding numbers of people in Kinshasa and Brazzaville. Nor 


should one overlook the role of popular music in spreading the language. 
Lingala is the language of 'Congolese music' — music which enjoys immense 
popularity throughout Zaire and Africa as a whole. 

A characteristic feature of Bantu languages is their system of 
alliterative concord. In this system, each verb agrees with its subject 
noun in class, number and person; and every variable modifier agrees 
with its head noun in the same way. Such agreement Is usually marked by 
prefixes. Spoken Lingala is unusual among Bantu languages since much of 
the system of alliterative concord found in other Bantu languages has been 
lost (as noted by Bokamba 1976, 1977; Ellington 1974, Guthrie 1951). 

In the following discussion, I examine words which 1) have qualities 
as their referents and 2) can be used to modifiy nouns. The word modifier 
is used as a cover term to refer to these words; the symbol FX] represents 

In some of the Bantu languages which have been analyzed as having a 
category adjective, the modification relation can be expressed syntactically 
by a noun immediately followed by a modifier which agrees in noun class with 
its head noun Cn XJ . The example below is from Swahili: 

1. mi-ti mi-kubwa 
N X 
trees big(ness) 
'big trees' 

Although it has been suggested (Ashton 1944, Welmers 1973) that the words 
appearing in the [xJ slot might be better categorized as nouns, they are 
still generally referred to as adjectives, primarily on the grounds that 
th^ immediately follow the nouns they modify and agree in noun class with 
their head nouns. 

In another Bantu language, Lonkundo, the modification relation can 
be expressed syntactically in the associateive construction Qn -a XJ, 
where the associative morpheme -a_ agrees in noun class with the head noun, 
but the modifier does not. The following example is taken from Welmers 




ho ns n£ 






'big trees 


Welmers analyzes the CXl in such constructions as nouns. 

InSpoken Lingala, the modification relation can be expressed by either 

[N x3 or [N -a X] , as shown below. 

3 a) nzete montni 3 b) nzete ya mon£nfc 
NX N assoc. X 

tree big(ness) tree big(ness) 

'a big tree' 'a big tree' 

Both of the above constructions (3a and 3b) have the same meaning and the 
same distribution. For every construction CN X] in Spoken Lingala, there 
is a corresponding construction [N -a Xj . While there is always a 
synonymous interpretation for the CN XJ and corresponding fN -a Xj , there 


may also be another interpretation for the [N -a Xj , derived from the more 
complex [n -a N Xl.'^ 

The associative construction is also open to the general class of nouns 
[N -a N]. 

4. moto ya mb^ngo 

N assoc N 
person money 
'a rich person' 

and to the general class of verbs [^N -a v3: 

5. likambo ya kokamwa 

N assoc V 
problem/ be amazed 

'a surprising affair/something amazing' 

In keeping with the analyses which categorize the [Xl in (3a) as an 
adjective (as in Swahili) and the Cx] in (3b) as a noun (as in Lonkundo) , 
traditional accounts of Lingala categorize mon£ nt as an adjective in (3a) 
but as a noun in (3b). So, for Spoken Lingala, it is not a matter of 
distinguishing between modifiers which occur in [. N Xj and those which occur 
in [n -a X] , since the same modifier may occur in both constructions. 
Rather, it is a question of determining how modifiers should be categorized. 

The characteristic Bantu system of agreement is shown in (6a) and (6b) 
below. These examples are taken from Standard Lingala (Bokamba 1977) and 
represent the general principle as it operates in Bantu languages.^ 

6 a) e-loko e^-ntnfc £^-kwey-aki awa 
thing big(ness) fall past here 
'a large object fell down here.' 

6 b) bi-lJ k3 bi.-nfcnt bi-kwey-aki awa 

things big(ness) fall past here 
'Large objects fell down here.' 

Note that in the above examples, the verb agrees with its subject noun 
and that the modifier agrees with its head noun, as indicated by the 
underlined prefixes. 

In Swahili, one of the criteria used in categorizing a word as an 
adjective is that it agree with its head noun, as in example (7) below. 

7 a) ki-su ki-refu 

N X 
knife long/length 
'a long knife' 

7 b) vi-su vi-refu 
N X 

knives long'length 
'long knives' 


In Lonkundo, on the other hand, one of the criteria used by Welmers 
in categorizing modifiers as nouns is that they usually do^ not agree with their 
head nouns, as shown in example (8) below. (But see note 2.) 

8 a) j-oi j-a i-sisi 
N assoc X 
thing small (ness) 
'a small object' 

8 b) ba-oi b-a to-sisi 
N assoc X 
things small (ness) 
'small objects' 

In Spoken Lingala, modifiers appear in both CN X] and Qn -a Xj 
constructions. However, the construction in which a modifier appears has 
no bearing on its agreement with its head noun. Aniong modifiers in Spoken 
Lingala, very few show any kind of agreement with head nouns; however, there 
is a small group which are inflected for number, on the model of the mo/mi (3/4) 
noun class. 

9 a) 

mwana mskj. 
bana mikt 
*bana bak£ 

9 b) moto monfcng 

bato mintnt 

*bato bantni 

9 c) 

moto molai 

bato milai 

«bato balai 

[n -a X] 
mwana ya mok€ 
bana ya mikt 
■•bana ya bakt 

moto ya montnt 

bato ya mintnt 

*bato ya bantnfc 

moto ya molai 

bato ya milai 

*bato ya balai 

a small child' 
small children* 

a big person' 
big people' 

a tall person' 
tall people' 

Most modifiers, however, never show any kind of agreement with their head 

10 a) mwasi kitikj 
basi kitakj 

10 b) 

mwana mabe 
bana mabe 

mwasi ya kitako 
basi ya kitoko 

mwana ya mabe 
bana ya mabe 

'a beautiful woman' 
'beautiful women' 

'a naughty child' 
'naughty children' 

The point here is that, in Spoken Lingala, agreement of modifiers and their 
head nouns is independent of the construction in which they occur. Therefore, 
analyses citing agreement of noun and modifier in the [N XJ construction as 
grounds for calling these modifiers adjectives will not work for Spoken Lingala. 

In Swahili, where the fact that modifiers agree with their head nouns 
is used to justify a category adjective, there are, in fact, some modifiers 
which never agree with their head nouns: 


11 a) ki-tabu safi 11 b) vi-tabu safi 


book clean(ness) books (clean(ness) 

'a clean book' 'clean books' 

Turning again to Lonkundo, where the status of modifiers as nouns 
rests partly on their lack of agreement with their head nouns, we find 
that some modifiers do in fact show noun class agreement with their head 
nouns : 

12 a) ba-nto _b-a ba-ljtsi 

N assoc X 
people good(ness) 
'good people' 

12 b) ba-na _b-a ba-be 
N assoc X 
children bad(ness) 
'naughtly children' 

12 c) ba-ya b^-a ba-tale 

N assoc X 

palm tall(ness) 

■^ trees 

'tall palm trees' 

As has been pointed out, agreement criteria brought to bear in making 
the distinction between adjectives and qualif icative nouns are not entirely 
convincing. There are enough exceptions, i.e., cases of agreement in the 
[n -a X2 construction (Lonkundo), and cases of non-agreement in the []N x] 
construction (Swahili) , that the analysis appealing to agreement offers no 
relevant insights into the data. 

It is also on the basis of syntactic distribution that it is claimed 
that modifiers are adjectives in Swahili, where they occur in the fN Xj 
construction, but nouns in Lonkundo, where they appear in the ^N -a x3 
construction. As shown in (3), modifiers in Spoken Lingala can appear in 
both constructions. On the basis of their distribution, one might conclude, as 
has been done in traditional analyses of Lingala, that modifiers in Spoken 
Lingala are both nouns and adjectives. However, in a theoretical framework 
where both categories are primes, it will clearly not do to have words which 
belong to both categories. The following distributional evidence is provided 
in support of the claim that all modifiers in Spoken Lingala are nouns, and 
that their adjective-like qualities would be better reflected by categorizing 
them qualif icative or adjectival nouns. 

Like nouns, modifiers can be the subjects of sentences and can also be 
further modified by possessives, as shown in the examples below where mbanga 
is a true noun' and mabe , a modifier: 

13 a) Mbongo ya ye eleki mingi. 
N poss. V ADV 
money his/her surpass much 
'She/he has a lot of money.' 


13 b) Mabe ya ye elekl mingi. 
X poss V ADV 
bad (ness) his/her surpass much 
'She/he is exceedingly bad.' 

In (14) molai which is a modifier, occurs in the same environment as mbjngj 
which is a 'true noun.' Both are the object of a preposition: 

14 a) Henri aleki Pierre na mbjngo. 
N V N Prep N 
Henry surpass Peter money 
'Henry has more money than Peter.' 

14 b) Henri aleki Pierre na molai. 

N V N Prep N 
Henry surpass Peter length/long, height/tall 
'Henry is taller than Peter.' 

The one construction in which there is a distributional distinction 
between quallf icative nouns [x] which can occur in [n Xj and other nouns 
is the 'have construction';, . 
[N have[^^j] 

where -zala na ('be with') means 'have.' In this construction, substitution 
of an [x\ for an [nJ results in either an uninterpretable sentence or one 
with a rather strange interpretation: 

15 a) Jean azali na mbongs. 

N V 

John have 

'John has money. ' 

15 b) Jean azali na 

N V 

John have 

'John has a wife.' 



woman/ wife 

16 a) *Jean azali na montnt . 
N V X 
John have big(ness) 

16 b) *Jean azali na mabe. 
N V X 
John have bad (ness) 

However, in example (17), it can be seen that in some cases, [x] can be 
Interpreted in the 'have construction', although the interpretation is a 
strange one or is attributed to a different construction altogether. 

17. Motuka ya ye ezali na mptmbfe . 
N poss V X 
car his/her have white(ness) 


The above sentence was interpreted by some Informants as meaning 'Her/his 
car is partly white.' (I.e., it has many colors, only one of which is white); 
while others interpreted it as meaning that there was something white (a can 
of paint or a piece of cloth, for example) in the car. 

The unacceptable nature of (16a) and (16b) and the interpretations of 
(17) may be analogous to the strangeness of (18b) below: 

18 a) Mwasi oyo azali molakisi. 
N Dem V N 
woman this be teacher 
'This woman is a teacher.' 

18 b) ? Mwasi oyo azali nzete. 

N Dem V N 
woman this be tree 
'This woman is a tree.' 

Both molakisi and nzete are clearly nouns, but only one of them is acceptable 
in the above sentence (ruling out, of course, metaphorical usage). This is 
further evidence that CX] is not of a different grammatical category than 
other nouns, but rather, is of a sub-category of nouns defined by the 
semantics of its referent. Any restrictions on its distribution in the 'have 
construction' are a result of its meaning. 

If modifiers in Spoken Lingala are nouns, why are they the only nouns 
which occur in the [N X] construction? As noted earlier, for every construc- 
tion [n xJ , there is a corresponding construction ^N -a XJ which can be 
interpreted as having the same meaning (and which may have another meaning 
as well) . 

19 a) ndako monEnfe 

N X 
house big(ness) 
'a big house' 

19 b) ndako ya mon^nt 

N assoc X 
house big(ness) 
'a big house' 

However, the [^N -a X] construction formed with non-qualif icative nouns in 
the |[x] slot and used to express 'adjectival' meanings does not have an \Ji XJ 
correspondent . 

20 a) moto ya mbongs b) * moto mbDngD 

N assoc N N N 

person money person money 

'a wealthy person' 

It would appear that qualif icative nouns are the only ones which can 
appear in the appositive Qn XJ construction. I would like to suggest that 
this is because the construction as a whole is subject to certain interpre- 
tation conditions — conditions which do not apply to the associative \Ji -a XJ 
construction. It would appear that what distinguishes the \N XJ from the 


LN -a XJ construction is the type of relationship between the [n] and the 
[X.1 appearing in them. The appositive construction [N x] may be reserved 
for those cases wherein the [X] can be interpreted as being an inherent or 
basic quality of its head noun. This may also explain the restriction on 
the distribution of qualif icative nouns in the ^ N have X] construction 
which one informant claimed is reserved for temporary attributes or 
possessions which can be relinquished by the possessor. 

In conclusion, modifiers traditionally categorized as both adjectives 
and nouns in Spoken Lingala can be more accurately categorized as nouns. 
In the [N x] construction, they do not agree with their head nouns and, 
morphologically, many of them behave like nouns of the 3/4 (mo/mi) class. 
The fact that their syntactic distribution differs from that of other nouns 
(i.e., they can occur in the [N Xj construction, but cannot — or rarely — 
occur in the lN have X] construction) is due to the semantics of their 
referents. Subcategorization of modifiers as qualif icative or adjectival 
nouns might allow us to predict which nouns could occur in the [^N XJ 
construction, but it is not necessary. Semantic interpretation conditions 
on the construction constitute a sufficient constraint. 

The behavior of modifiers in Spoken Lingala is noun-like. This can 
best be accounted for by a theory which provides a set of syntactic 
distinctive features defining the possible lexical categories of human 
languages (see Jackendoff 1977:29). Such a theory is capable of capturing 
the 'adjective-like' semantic quality of modifiers in languages where they 
behave like nouns as well as in those languages where they behave like 
verbs . 


1 This paper was prepared for the 10th Annual Conference on African 
Linguistics, held at the University of Illinois in the Spring of 1979. 

Lonkundo data in this paper are from Welmers (1973:273,4). According 

to one of the reviewers of this paper, the following fN XJ construction is 

also acceptable, and can serve as subject and as predicate nominal: 

be-tamba be-nene. 

■^The data from Spoken Lingala come from informants. I am grateful to 
the following people who have provided me with data and offered insightful 
comments on their language: Salikoko Mufwene, Lukaya Ntoya, Constance 
Ntoya, Bellamy Nignon, and Annette Onema Diawara. 

^An example of this is the phrase mwasi ya kitokj , which may be inter- 
preted as meaning either mwasi klto k3 or mwasi ya nzoto ('body') kitskj . 
The former refers to personal character/behavior, physical beauty, etc., 
while the latter refers only to physical beauty. 

^Standard Lingala is the formal variety of Lingala and is taught in the 
schools, is used on the national and armed forces radios. Compared to 
Spoken Lingala, the use of Standard Lingala is extremely limited. 




ASHTON, E.O. 1944. Swahili grammar: Including intonation. London: 

BOKAMBA, Eyamba G. 1977. Ekolongonelo ya Lingala: an introductory 

course. Urbana, Illinois: University of Illinois. 
. 1976. Question formation in some Bantu languages. Indiana 

University Doctoral Dissertation. 
BRYAN, M. A. 1959. The Bantu languages of Africa. London: Oxford 

University Press. 
BWANTSA-KAFUNGU. 1972. J'apprends le Lingala tout seul en trois mois. 

Kinshasa: Centre de Recherches Pedagogiques. 
. 1970. Esquisse grammaticale de Lingala. Kinshasa: Publi- 
cations de I'Universite de Lovaniura. 
ELLINGTON, John. 1974. A handbook of Kinshasa Lingala. 
GUTHRIE, Malcom. 1967. Comparative Bantu. Farnborough, Hants: Gregg 

International Publishers. 
. 1951. Grammaire et dictionnaire de Lingala. Kinshasa: 

Conseil Protestant du Congo (published by Gregg International). 
HAGEGE, Claude. 1974. The 'adjective' in some African languages. Studies 

in African Linguistics, Supplement 5, October 1974. 
JACKENDOFF, Ray. 1977 "X syntax. Linguistic Inquiry Monograph //2. 
MISSION EVANGELIQUE DE I'UBANGI. c. 1950. Lingala grammar: lessons and 

exercises. Kinshasa: CEDI . 
SESEP N'SIAL BAL-A-NSIEN. 1978. Le metissage Franjais-Lingala au Zaire: 

Essai d'analyze dif f erentielle et socio-linguistique de la communication 

bilingue. Universite de Nice these de doctorat. 
VAN EVERBROECK, R. 1969. Le Lingala parld" et ecrit. 
WELMERS, William. 1973. African language structures. Los Angeles: U.C.L.A. 




Studies in the Linguistic Sciences 
Volume 14, Number 2, Fall 1984 

Eluzai M. Yokwe 

The aim of this paper is to examine the indigenization (Ara- 
bicization) of language policy as opposed to the European language 
policy in the Sudan. The political, socio-cultural and linguistic 
aspects of the country are presented. The use of Arabic as a lin- 
gua franca in relation to the use of the other indigenous languages 
throughout the country is demonstrated, while the concept 'Arabici- 
zation' is analysed in its socio-political and linguistic perspec- 
tives. Against this background, language policy which is based on 
the general policy of Arabicization to the exclusion of the indige- 
nous languages is discussed. It will be shown that the Sudan is 
deviating from the comprehensive language policy to adopt a totali- 
tarian policy that favors one out of many ethnic groups of the same 
country. It is argued, in this paper, that there is every reason 
for the Sudan in particular and Africa in general to adopt a com- 
prehensive language policy which assigns appropriate functions to 
each lingua franca and the vernaculars wherever the need arises. 
The paper concludes with strong support and an appeal for the re- 
turn to a comprehensive language policy that is both realistic and 
flexible . 


There is a growing awareness among the African countries for the need to 
indigenize their language policies as opposed to the European language poli- 
cies and to adopt a comprehensive language policy which assigns appropriate 
functions to each lingua franca and the vernaculars. Such a language policy 
is critical to the African system of education and cultural development (Bo- 
kamba & Tlou 1976; Kalema et al. 1980). 

Guided by comprehensive language policies, Tanzania, Somalia and Ethio- 
pia have responded to this need by promoting and adopting Swahili, Somali and 
Amharic languages, respectively, into official languages and by using them as 
media of instruction in schools up to a certain level. In the Sudan, parti- 
cularly in Northern Sudan, this indigenization policy has meant total Arabi- 
cization of the school system, the mass media and the public communication 
network to the exclusion of the other native languages of the country. The 
Government has been leaning toward the adoption of a totalitarian language 
policy based on the great policy of Arabicization, which simply means socio- 
cultural and linguistic assimilation of the indigenous groups. In addition 
to this, Arabicization has been emotionally associated with the political 
and socio-cultural goal of the country. For example, a former minister of 
interior, Ali Abdel Rahman, retorted one time in his policy speech concern- 
ing the goal of the country, by saying: 


The Sudan is an integral part of the Arab World 
... Anybody dissenting from this view must quit 
the country. ' 

Such expressions are commonplace political rhetoric as I will show in 
this paper. The Arabicization policy in the Sudan is being popularized by 
the central government in Khartoum against the other ethnic cultural groups. 
As a result, the other Sudanese communities such as the Southerners ^ are 
left without any option except to fight back. Thus, in 1963 a petition was 
submitted to the United Nations by a group calling itself the Sudan African 
Closed Districts National Union which stated that: 

Vernacular languages which vn-re taught at lower 
schools were replaced by Arabic to be taught with 
"Koran", the Muslim Bible. Village schools where 
elementary subjects were taught have been closed 
down and replaced by "Khalwas" (Islamic schools) 
... The slightest mistake done by a Christian pu- 
pil who is not attending Islamic lessons is enough 
to dismiss him from the school. Children in the 
South are made to study their lessons in Arabic at 
the early stages of education because the idea is 
to make Arabic the mother tongue. This we think 
is wrong according to educational psychology prin- 
ciples. The child should begin with his native 
language not a foreign one. 

The argument of this paper is that there is every reason for the Sudan 
in particular and Africa in general to adopt a comprehensive language policy 
which assigns appropriate functions to each lingua franca and vernaculars 
wherever the need arises. 


The Republic of the Sudan constitutes 967,000 square miles (the largest 
in Africa) and is divided into nine administrative regions (See Appendix A, 
Map); six in the Northern part and three in the Southern part of the country. 
The total population of the Sudan is about 18 million (World Bank, 1980). Of 
this number, six out of 18 million people are Southerners, while 12 million 
are Northerners. Two predominant racial groups occupy the Sudan - the Arab- 
ized Sudanese in the North and the Negroid people in the South. Ninety-five 
percent of the Northerners have become Muslims and have been assimilated into 
the Arab culture; the remaining five percent are Christians, who are also 
greatly influenced by the Arab culture. In the South, ninety-six percent of 
the population are Christians, one percent Muslim, and the remaining three 
percent other. 

There are 136 languages spoken in the Sudan; 114 are Sudanese, 14 origi- 
nate from other African countries and eight, including English, are European 
languages (Bell 1975; Thelwall 1978). Fifty-one percent of the Sudanese peo- 


pie speak Arabic as their mother tongue; twenty percent of the population 
speak Arabic as their second language. In the South, however, only five 
percent of the Southerners speak Arabic (Southern Arabic) as their second 
language. Dinka, spoken by eleven percent of the Sudanese population, is 
the second major language in the country. Fourteen Sudanese languages, all 
classified as minor languages, are spoken by thirty-one percent of the to- 
tal population. Although English is spoken by only one percent of the Su- 
danese population, this number represents a very influential segment, e.g., 
government officials, professionals, administrators, lecturers, etc. 

Socio-economically and politically, the Southerners, who are by and 
large Christians, treat themselves as distinct from the Arabized Northern- 
ers. The distinction was understandably recognized by the British Colonial 
Administration and became embedded in the 'Southern Policy of 1930' ** based 
on the following two premises: 

1 . That the Negroid Africans of the South were cul- 
turally and, to some extent, racially distinct 
from the Northern Arab Sudanese. 

2. That the Southern provinces would either develop 
eventually as a separate territorial and politi- 
cal entity or be integrated into what was then 
British East Africa. 

It was this attitude which was encouraged by the British colonialists 
that was predominant before the national reconciliation (1972), when the 
Southerners were granted local autonomy under one united Sudan. This North- 
South division between the Sudanese people is well summarized by Jaden 
(1965): 5 

The Sudan falls sharply into two distinct areas, 
both in geographical area and ethnic group, and 
cultural systems. The Northern Sudan is occupied 
by a hybrid Arab race who are united by their com- 
mon language, common culture, and common religion; 
and they look to the Arab world for their cultural 
and political inspiration. The people of the Sou- 
thern Sudan, on the other hand, belong to the Afri- 
can ethnic group of East Africa. They do not only 
differ from the hybrid Arab race in origin, arran- 
gement and basic systems, but in all conceivable 
purposes .... 


The use of Arabic relative to the other Sudanese and foreign languages 
has been placed on top priority for investigation in language surveys con- 
ducted by the Sudanese Government. The Institute of African and Asian Stud- 


ies (lAA), University of Khartoum, among others, has been particularly pre- 
occupied with the role of Arabic in the Sudan. Language studies and socio- 
linguistic surveys have been carried out by many individual linguists such 
as Stevenson (1962), Thelwall (1971), Mahmud (1983) and others. All point 
to a similar conclusion, that although Arabic is predominantly used as a 
lingua franca in the Sudan, it nevertheless is not and never will be a sub- 
stitute for the vernaculars which are still serving very important communi- 
cative roles, especially among the families in the rural areas of the Sudan. 
For example, vernaculars are still the recognized vehicles of transmission 
for folklore, traditions and local beliefs for the vast majority of the Su- 
danese people. 

One of the major language surveys of the Sudan was undertaken by Jernudd 
(1972) in the six provinces of Northern Sudan. This survey also covered one 
school area in the town of Adong, a small town in the Upper Nile Province in 
Southern Sudan. In response to a question formulated to assess the parents' 
(both fathers and mothers) ability to "chat along" in Arabic and the ability 
to read and write, it was discovered that 90% of the fathers in Heiban, the 
southernmost point visited in Northern Sudan, claimed knowledge of Arabic. 
By contrast, only 40% are said to know Arabic in Adong, while the results 
from the rest of the areas visited in Northern Sudan showed that almost all 
of the fathers knew or spoke Arabic. The mothers' knowledge of Arabic var- 
ies a little from one area to another; it is at its lowest in Adong (where 
approximately one quarter of the mothers are said to know Arabic), and the 
highest in the rest of the areas visited in Northern Sudan. 

With regard to the use of Arabic and vernaculars in the homes, markets, 
and offices, the survey pointed out that Arabic predominates in the markets, 
schools and offices, but within the realm of home and family, the vernacular 
languages predominate. In the Nuba Mountains (Heiban) and in Adong, the 
younger people are influenced by school, and tend to use more Arabic than 
the older people, particularly in the villages. The survey also found that 
these young people, contrary to expectation, do still use vernacular langu- 
ages just as much as the rest of the people in family settings and similar 
contexts: With regard to this finding, Jernudd (1979:63), commented that: 

It should imply that Arabic is functioning well as 
a lingua franca in at least Northern Sudan . . . the 
lesser influence of Arabic in the Southern part of 
the Nuba Mountains . . . may be reflected with a very 
slight margin through the lower Arabic figures for 
Heiban. The much lower figure for Adong implies 
that Arabic may only be used for limited communica- 
tion with people who otherwise do not share langu- 
ages . 

Another important language survey of the Sudan is that of Professor 
Thelwall (1978). Given below is a chart from his work (1978:4) in which 
he graded the provinces in terms of their degree of linguistic fragmenta- 




Provinces in the North 

Provinces in the South 



96.0 % 



84.0 % 

W. African 

8.0 % 



81.0 % 


19.0 % 



50.0 % 


36.3 % 

W. African 

11.0 % 



55.0 % 


21.0 % 


12.5 % 


5.0 % 



68.0 % 


26.8 % 


9.0 % 



68.0 % 


A. 6 % 

Ndogo Sere 

2.8 % 


0.9 % 



52.0 % 


25.0 % 


13.7 % 


4.8 Z 


1.7 % 



23.8 % 


19.4 % 


18.4 % 


13.0 % 


10.0 % 


4.0 % 


3.0 % 


0.6 % 

Thelwall's chart above demonstrates a wider role for the local vernacu- 
lar languages throughout the country. In the Northern Province, Arabic is 
spoken by 81% of the population and Nubian is spoken by 19%; while in Kassala 
province, Arabic is spoken by 36.3% of the population, Beja is spoken by 50%, 
and Hausa and Fulani (from West Africa) are spoken by 11%. These figures 
show that: 

The numerical superiority of Beja at the province 
level contradicts the popular idea that multilin- 
gualism in the Sudan is restricted to the three 
Southern provinces and that elsewhere it is neg- 
ligible (Thelwall 1978:8). 

Not only is Beja holding the 'numerical superiority' at the provincial level, 
but it is also the vehicle of the folklore, oral traditions and poetry of the 
Beja people. Such an important linguistic role seems to constitute a cultur- 
al strength that is not directly affected or revealed by literacy rates 


(Andrejewski 1968; Adarob 1972). Likewise, the "maintenance of Nubian as a 
first language among the Nubian peoples, is a proof of its cultural vitality 
in the face of a very long history of culture contact with Egyptian and la- 
ter Arabic" (Shinnie 1978:82). 

In the three Southern provinces, the use of Arabic is very insignifi- 
cant with an average currency of 0.73% in each province. This percentage 
necessarily represents the town populations in the South who use Arabic as 
a lingua franca in context situations other than home. In the evaluation 
report of Southern Sudanese local languages (Cowan & Cziko 1984), the sur- 
vey recorded that 93% of those interviewed felt most comfortable speaking 
their own vernacular languages, "but because of the heterogeneity of lan- 
guage usage in Southern Sudan, they sound it necessary to speak at least 
one or more additional languages." For example. 

If a Zande speaker is speaking to a person who 
speaks a vernacular he himself does not speak, 
or Arabic, they will use Arabic or another com- 
mon language which they can both use and under- 
stand (Cowan & Cziko 1984:12). 

The type of Arabic spoken in the South is, however, referred to as a 
pidgin-creole or Southern Arabic (Tucker 1934; Agwo 1975; Mahmud 1983; Yokwe 
1984). In my article, "The Diversity of Juba Arabic" (Yokwe 1984), which 
was mainly descriptive of the grammatical aspects of Juba Arabic, the follow- 
ing observation was offered: 

Arabic is used invariably in situational con- 
texts where one is not sure of the other's 
linguistic background. Thus, Arabic is taken 
as a lingua franca or assumed so, so that it 
is safer to talk to someone you do not know 
well in Arabic. 

It must be emphasized here that this pattern of language use (Arabic- 
Vernacular) is only true within the town populations in the Southern Sudan. 
Outside the towns, in the rural areas, the vernaculars are predominantly 
used. In fact, Arabic is almost non-existent in the rural areas where the 
majority of the Southerners live. It is this linguistic situation, then 
which led the regional government to formulate its language policy for educa- 
tion in such a way that the major vernaculars are used as media of instruc- 
tion within the first four years of primary schools in the rural areas (See 
Appendix B) . 

However, in the Northern Sudan, Arabic has been assigned the role of 
medium of instruction in all the pre-university schools. At the secondary 
school level, English is introduced as the second language but taught as a 
subject. There is no educational role assigned to the vernacular languages 
in the Northern Sudan; the reason being that the Arabicization language pol- 
icy as practiced in the North, neither allows the teaching of vernaculars in 



the schools, nor does it encourage their development. The assumption is that 
it is a waste of time to teach a child in his vernacular for a few years and 
then shift to a foreign language during the later stages of his school life. 
More will be said on this aspect in the coming sections. Before proceding 
to that discussion, let us first examine the concept of Arabicization and its 
logical implications as practiced in the Sudan. 


The term 'Arabicization' refers to several concepts. At one time, it 
referred to the process of racial, religious and cultural assimilation of 
the indigenous ethnic groups of the Northern Sudan; for example, the Nubians, 
Beja and other Negroid peoples by the Arabs. This process led to the Islami- 
zation and Arabicization of the six Northern provinces, leaving the South vir- 
tually untouched by these influences (Muddathir 1968:230). Thus, Islamization 
and Arabicization gave the Nubians, Beja and other Negroid populations of the 
Northern Sudan a unifying cultural bond which they did not possess before 
(Muddathir 1968). This view is widely held by the Northern Sudanese scholars 
who insist that "Arabism is a cultural, linguistic and non-racial link that 
binds together numerous races: black, white, and brown" (Beshir 1968; Mudda- 
thir 1968; Mazrui 1972). 

The second concept of Arabicization is a political one. It is a deliber- 
ate attempt by the Northern politicians to identify and shape the future goal 
of the country toward Arab nationalism instead of African nationalism. As 
early as the 1930s * it was declared by one of the Northern leaders speaking 
for his party, that: 

Sudanese nationalism must be firmly based on 
Islam, Arabic culture, and African soil and 
traditions and that it should be opened to 
and freely interact with, international cur- 
rents of thought. 

It remains to be know how this conviction was arrived at. However, si- 
milar politically motivated proclamations started to follow even after the 
Sudan became independent (1956). In 1965, for example, Sayed Saddiq el-Mahdi, 
a leader of the Umma Party (one of the Islamic sectors), characterized the 
Sudanese national image as follows: 

The dominant feature of our nation is an Is- 
lamic one and its overpowering expression is 
Arab, and this nation will have entity iden- 
tified and its prestige and pride preserved 
except under an Islamic revival. ^ 

Many others, such as Dr. Hassan Turabi (Wai 1973), leader of the then Islamic 
Charter Front (another Islamic sector) had on several occasions expressed him- 
self in a similar fashion. He argued that the South had no culture, so this 
vacuum would necessarily be filled by Arab culture under an Islamic revival. 


This view dominated the discussion of the first and second constitution com- 
missions in 1967 and 1968. One could go on and on enumerating these socio- 
cultural and political conceptualizations of Arabicizat ion by the Northern- 
ers. Interestingly, some claims have become too ambitious and desperate in 
their outlook. In one of his proclamations, Saddiq asserted that "The South 
is a stepboard for Arab entry and Islamic influence into the heart of Africa" 
(Vigilant, 1/16/1966,3). I perscnally find this overambit ious in that most 
of the African countries South of the Sudan or South of the Sahara, with the 
exception of a few in West Africa, are nations that are mainly Christian in 
religion and African in culture. It would be very difficult under the pre- 
sent circumstances for such goals to be realized, particularly when the image 
of Africanness is being subordinated to that of Arabdom. 

The desperate state arises when the same person iSaddiq) in Vigilant 
(1/16/1966,2), accused East Africa of conspiring against Islam and Arabism: 

Islam should influence the whole of Africa but 
'there is a conspiracy in East Africa.' Here 
people believe they are Negroes, different from 
Arabs and must project their own personality 
and follow their own way. This belief underlies 
the affairs in the Southern Sudan at present. 

Such a state of affairs challenges and contradicts the African spirit of 
struggle against colonialism and foreign domination of the African states. 

The third view of Arabicization policy is the linguistic role of the 
Arabic language as a means for achieving national integration and unity of 
the Sudanese people. ^The supporters of this view, which is widely held in 
the Third VJorld, believe that unity and national integration can easily be 
promoted by the adoption and the usage of one national language. In the 
case of the Sudan, obviously Arabic happens to be the language in question. 
It is not only that Arabic commands the majority of the Sudanese people in 
its role as a lingua franca, but that it actually complements the policy of 
Arabicization. The Arabic language, being the mother tongue of the Arabs 
in the Sudan is, therefore, bound to affect the outlook of the self- identi- 
fication of the Sudanese nation as a whole. > 

Such is the conceptualization of the Arabicization policy upon which 
language policy is based in the Sudan. Thus, it is not a coincidence that 
the national language policy of the Sudan provides no room for the indige- 
nous languages, particularly in the North where Arabic is mostly used as 
an official language and as a lingua franca as well. It appears as if the 
Arabicized Northerners themselves have given up the idea of developing and 
preserving their mother tongues. The pressure from the national government 
against traces of African culture, traditions and languages is overbearingly 
too much. Indeed, the same pressure is now being exerted toward the South 
which has been resisting the policy of Arabicization ever since the nine- 
teenth century. This is not because English is preferred to Arabic by the 
Southerners, but because of the vehement socio-cultural and political strings 
which the Arabic language carries with it as it has been demonstrated above. 


Thus, the Northern political outcry to denounce the African face of the 
country, in favour of the Arab one, alerts the Southerners to the danger 
that threatens their socio-cultural and linguistic heritage. As a result, 
the Southerners have always rejected Arabicization policy of any form, in- 
cluding linguistic form. They, however, support a language policy that is 
comprehensive enough to include the vernaculars in its structure and yet 
promote Arabic as a national language. 


5. 1 Language Policy in the North . The existing language policy of 
the Sudan was decided on the basis of Arabicization policy: the pursuit of 
"Arab Nationalism" as the future goal of the Sudan. It is a national unity 
and integration sought in the "sentimental uniformities ... based on socio- 
cultural integration" (Fishman 197 1:5). 'Arab Nationalism' was being adop- 
ted and transformed into the national language policy by the government, 
much to the exclusion of the rest of the languages and cultures of the in- 
digenous people. In the North, it was easily done since the socio-cultural 
integration had already been achieved in the 17th and 18th centuries. The 
adoption of the Arabicization of the language policy was first declared on 
the eve of independence ( 1953) by the first Sudanese Minister of Education. 
Speaking in the National Assembly (Wai 1973:18) on behalf of his government 
he asserted that: 

As the Sudan is one country sharing one set of 
political institutions, it is of great importance 
that there should be one language which is under- 
stood by all its citizens. That language could 
only be Arabic, and Arabic must therefore be 
taught in all our schools. 

Before the implementation of this policy in 1958, an event occurred which 
gave a sweeping mandate to the government to pursue its language policy. 
The policy was supported by a favourable recommendation on the use of Ara- 
bic as the medium of instruction in the secondary schools of the Sudan by 
the International Education Commission on Secondary Education (Commission 
Report, 1955). Among other things, the Report of the Commission stood 
against the use of vernaculars as media of instruction in schools and 
recommended Arabic to replace them. One of the reasons for recommendation 
was stated thus: 

It would be a waste of time and energy to try to 
teach the children of the South in their own Ver- 
naculars in which they will not be able to pursue 
any reading after they leave school: such Verna- 
culars have no literature and cannot be used as 
cultural media (Commission Report, 1955). 

This particular version of recommendation became the most celebrated and 
most often quoted statement by the Khartoum governments and a number of 
scholars whenever the issue of language policy was discussed in the Sudan. 


As late as 1963, a UNESCO experz, for example, re-echoed this allegation to 
a conference on Adult Education in the Sudan. He asserted that: 

The tribal languages of the South have no script 
of their own and even if the Latin alphabet is 
used for the different languages, there is no 
literature worth speaking of, which can be read 
with pleasure and profit. * 

Such sweeping statements are indeed uncalled for, particularly when none of 
the speakers had knowledge of any Sudanese languages, given the fact that 
the Commission only spent six weeks in the Sudan including a few days in the 
South. Such an allegation was indeed the more lamentable particularly when 
the members of the Commission never bothered to find out for themselves what 
was going on in that area regarding the language situation (Sanderson et al. 
1981). Indeed, as absurd as the report was, two of the members of the Com- 
mission, Sir Charles Morris and Miss Charlesworth , registered their precau- 
tions which the Sudanese Secretary ^ then excluded from the main body of the 
report. It appeared later in another document entitled "Summary of the Main 
Conclusions and Recommendations" (1955:144-5). Their suggestion was quite 
realistic and quite anticipatory of the later ideas of language policy con- 
cerning the South. The two recommended that where the vernacular is exten- 
sive and where there exists some literature and strong local sentiment, the 
vernacular should continue as the language of instruction in the earlier 
stages of primary education (The Commission Report, 1955). 

As a result of the Arabicization policy, the use of Arabic was intensi- 
fied in government institutions and public sectors, particularly in the North. 
Arabic is the language of religion used for prayers in the mosques and Chris- 
tian churches. The Coptic Church holds services in Classical Arabic as spoken 
in Egypt. The Roman Catholic Church uses Colloquial Sudanese Arabic for its 
vernacular liturgy, and uses it throughout the Northern Sudan. In mixed eth- 
nic congregations, the Anglican Churches use Colloquial Sudanese Arabic, but 
in single ethnic congregations, the relevant vernacular is used (Thelwall 
1978). Most of the newspapers are in Arabic and three-quarters of the time 
of Radio Omdurman (the national radio station) is spent on broadcasting in 
Arabic. The Arabic language is also used on television, except when imported 
foreign language films are shown. 

From 1958 onward, Arabic became the medium of instruction from primary 
schools up to the secondary level in the North. English is taught as a sub- 
ject from intermediate school onward, and it remains the most common second 
language. It may be mentioned here that the Arabicization language policy 
was later reinforced by the permanent constitution of the Sudan (1973), writ- 
ten in both Arabic and English, which among other things stated that: 

Arabic shall be the official language for the 
Sudan and English the principal language for 
the Southern Region without prejudice to the 
use of any language or languages which may 


serve as a practical necessity for the effi- 
cient and expeditious discharge of executive 
and administrative function of the Region. 

Note that this constitution is one of the brakes against absolute Arabici- 
zation policy, including language policy. 

The constitution, properly referred to as 'The Southern Self-Govern- 
ment Act (1972)' or 'Addis Ababa Agreement (1972)', resulted from the 17 
years of civil war that took place between the Southerners and the North- 
ern governments from 1955-1972. This political document is embedded 
in the national constitution as an organic law that allows the South to 
form an Autonomous Regional Government within the United Sudan. As a 
matter of fact, the central government has no language policy in the real 
sense of assigning to language or varieties of a language social functions 
within the respective communities. What the government has done is simply 
to state the obvious: the Sudan is an Arab country and so Arabic is the 
language for every function. No consideration is given to the other Afri- 
can languages spoken in the North. 

5.2 Arabicization Policy in the South . Arabicization in terms of 
national integration by the process of socio-cultural assimilation of the 
Southerners has had very little success. The Southerners, just like their 
counterparts in the North, still remain a socially distinct community. 
The socio-cultural and linguistic assimilative revolution that swept across 
the North Sudan in the 17th and 18th centuries seemed to have lost its mo- 
mentum toward the South. Nobody has come up with a definite answer based 
on solid research as to why this gap between Northerners and Southerners 
is still so wide and deep. From my own point of view, I think that the 
most deep-rooted problem between the two sides is the lack of 'trust' in 
one another. There are several factors which have contributed to this dis- 
trust, including the bitter history of slavery. The most important factor, 
however, is the lack of respect, by Northerners, for the Southerners' 
rights, culture, and beliefs. 

Specifically, in some communities self-dignity and respect are easily 
and readily sacrificed for bread and butter or for political position or 
money. In the South and among the majority of the various ethnic groups, 
the reverse is true: self-dignity and respect come first and other things 
follow later. Our Northern friends who claim the Islamic civilization seem 
to behave contrary to the expectations of the Southerners. They expect the 
Southerners to readily give up their cultural and traditional values in ex- 
change for Arabdom and when the response is not forthcoming, they resort to 
either force or hatred. As Beshir (1968:80) notes, "when the parliamentary 
system disappeared, and political parties were suppressed, the advocates of 
compulsion and integration of the North and South by force of arms had the 
upper hand." 

The military re^.ime of 1958, which most Southern officials of that 
time considered as a conspiracy by Northerners, was a monopoly of the Arab 
North. Although political opposition was suppressed on either side, the 
military regime set out to carry the policy of Arabicization on full scale 
in the South. According to Beshir (1968:81), the military regime: 


Stepped up the spread of Arabic and Islamization, 
in the belief that this was the only way to ach- 
ieve unity in the future. A number of Koranic 
schools were established in different districts 
and Islamic priests were appointed. Six inter- 
mediate Islamic institutes were opened in Juba, 
Kodak, Wau, Mar id i, Yei, and Raga . A secondary 
Islamic institute was opened in Juba, and centers 
for preaching and religious instruction for adults 
were also established. The military governors and 
administrators devoted much of their time and en- 
ergy to spreading Arabic and Islam and to suppress- 
ing the opposition. 

This prolific attempt to assimilate the South was done under the pretext 
that it was the only way to achieve unity in the future. Such a kind of 
unity was doomed to failure because there was nothing to unite. The Sou- 
therners, most of whom were already Christians, found themselves faced 
with a "clear and conscious choice between two rival religious, ethnical 
and cultural systems" (Sanderson et al. 1981:394). 

However, a quick glance at Tables II, III and IV below and on the 
next page, will show that the other government institutions, including 
the schools, police, and army, remained a monopoly of the Northerners 
(Alibino 1970). 



No. of 
in North 

No. of 
in North 

No. of 
in South 

No. of 
in South 



Added after 






*Malakal Secondary School, which bears the name of a town in the 
South which was opened in Omdurman in 1962, has now been absorbed 
into Wadi Seidna Secondary School. 





No. of 

No. of 


































Total 190 21 211 

*There was no intake in 1952, 1954, 1955, 1956, 1958 or 1959. 



No. of 

No. of 





27.7. 1954 








4. 1956 




7. 1957 




5. 1958 




5. 1959 








1. 1962 




1. 1963 












Total as at 

1. 1.1965 




While the statistics in Tables II, III and IV are twenty to thirty years 
old, very little has changed in this pattern of development: the Northern 
part of the country continues to be the target of attention. The figures 


in these tables indicate that the government policy has certainly been fa- 
vouring the Northern region. 

The attitude of the North toward the South is a mixture of hatred and 
disrespect, and this is the biggest factor that has contributed to the fail- 
ure of the policy of Arabicization in the South. This view is supported in 
several studies in the literature. Henderson (1965:153) states in this re- 
gard, for example, that: 

One thing they shared was a common contempt 
for the Southerner as an inferior being, 
coupled with complete indeference to his re- 
ligious ideas, ethics and standard behaviour, 
his social and tribal pattern. 

Another interesting observation was indeed made as part of the 'Report 
of the Commission of Inquiry (1956)' '' formed by the government. The Com- 
mission states, in one of its sections that: 

It is unfortunately true that many Northern 
Sudanese especially from among the educated 
class, regard the Southerners as of an infer- 
ior race, and the Gallaba (Northern traders) 
in Southern Sudan form no exception to this, 
as the majority of them are uneducated. The 
traders refer to the Southerners, and often 
call them 'Abeed' (slaves). This practice 
of calling Southerners 'Abeed' is widespread 
throughout the three Southern provinces. It 
is certainly a contemptious term, and is a 
constant reminder to the Southerners of the 
old days of the slave trade. 

On the other hand, many Southerners also seem to be contemptious of North- 
erners, as is shown by the terms Mundukuru and Minga which they occasionally 
use to refer to them. The two terms are nicknames given to the Northerners 
by the Southerners, in reference to the Arab traders in the South whose in- 
terest in making money is so great that they subject themselves to such sim- 
ple food as kisera 'Sudanese pancake' to be taken with water and salt. Over 
the course of time, however, th'ise words were used for demeaning any North- 
erner, particularly when a situation is provoked. Apparently the Southerners 
have learned the fact that the Arabized Northerners like to be called Arabs 
despite the fact that they are not. To deny the Northerners that claim, the 
terms Mundukuru or Minga are substituted for the term 'Arab'. When used in 
this context, the terms mean 'Arab mulatos' or 'Arab half-castes'. 

The end result of all of this bitterness was the seventeen years of 
civil war ( 1955-72) that brought about the formation of the Regional Govern- 
ment within the United Sudan. And with the Addis Ababa Agreement (1972), the 
old chapter was closed and a new one opened. What it really meant was that 
unlike the situation in the North, the socio-cultural policy of Arabicization 


had failed to a large degree in the South. However, linguistically, the role 
of Arabic as a national language has been preserved. The Arabic language is 
not new to the people of the South: it has been known to the South ever 
since the Anglo-Egyptian rule. Tucker (1934) commented on "Southern Arabic" 
as being widely spread throughout the Southern Sudan. The colonial govern- 
ment (Southern Policy 1930:4) ruled that English instead of Arabic be used 
where communication in the local vernacular was impossible. 

Every effort should be made to make English the 
means of communication among the men themselves 
to the complete exclusion of Arabic. 

The Rejaf Language Conference (1928) which laid the foundation for the devel- 
opment of the local vernaculars and English in the Southern schools, recom- 
mended that Arabic in Roman script also be required in certain communities 
where the use of no other vernacular is practicable. 

5.3 Language Policy in the South . The language policy in the South, 
during and atter inaepenbence, was Dased on the idea that the native children 
be taught how to read and write in their native languages. After that, they 
will be gradually introduced to English or Arabic (in third or fourth years 
of primary school). Much of the educational task was left in the hands of 
the missionaries who were greatly in favour of the use of the vernacular lan- 
guages in the first three years of primary school learning. 

In 1972, after the Addis Ababa Agreement, the regional government formu- 
lated a language policy that was quite comprehensive, at least in my opinion. 
Accordingly, it was resolved (Yokwe 1977:1; Mahmud 1983:156) that: 

The High Executive Council hereby endorsed the 
use of local languages and of Arabic and English 
for education in the Southern Region. (See Appen- 
dix B) 

In my opinion, this is a comprehensive language policy as opposed to the to- 
talitarian language policy of Arabicization. In rural schools, the local 
vernacular is the language of instruction for the first four years of school; 
Arabic and English are introduced orally. In urban schools, Arabic is the 
language of instruction through the sixth year of primary school. '^ with 
English being introduced orally. (See Appendix B for details) Unlike the 
Arabicization policy, this language policy reflects and accommodates the 
realistic and pluralistic multilingual societies of the Sudanese nation. As 
Cowan (1984:76) clearly stated: 

The language policy eventually adopted by the 
Southern Regional Government represents an in- 
teresting compromise designed to accommodate 
the central government's insistence on main- 
taining Arabic in the educational scheme of 
the Southern Region while insuring a favored 
status to the South 's "principal" language 


English, and, at the same time, allowing for 
mult il ingualism. 

Above all, the Southern language policy is comprehensive, flexible 
and open for, not only "wide discussions" but open for future amendments 
to serve practical purposes. One example of practical purpose is the ac- 
commodation of Arabic as a medium of instruction in the Southern schools, 
particularly in the towns. It was not because the Southerners appreciate 
the Arabicization policy from the North, nor is it that Arabic, of late, 
is selling hot to the Southerners; it was simply because the Southerners 
were faced with a real linguistic problem in the schools whose pupils 
come from a diverse linguistic background living in the same town or city. 
When a situation like this one arises, the Southerners have no option ex- 
cept to turn to Arabic which is also the national language and the lingua 
franca in the cities. The truth is that Southerners do not treat verna- 
culars or English as substitutes for Arabic. On the contrary, they want 
a policy that is comprehensively authentic enough to be called Sudanese 
which does not necessarily mean Arabism nor Africanism. The governments 
in Khartoum should aim at a policy that reflects our identity as Afro-Arab. 

Unlike the Arabicization policy, the Southern language policy is em- 
pirically supported by the results of the various recent research regarding 
the importance of the use of mother tongue in teaching children how to read 
and write in their first three to four years of primary school. 

It has been shown (Resnick 1968; Bamgbose 1984; Fishman 1984) that 
children are better able to master the basic skills in reading, writing and 
arithmetic when these are introduced in the child's own language (or at 
least in the language of the community in which he or she resides). Nor 
does initial schooling in mother tongue interfere with the child's later 
acquisition of another language — which has been one of the major arguments 
of proponents of mother tongue instruction. On the contrary, studies have 
shown that children who have a firm grounding in their own language are bet- 
ter able to master a second language. Numerous studies in various parts 
of the world support this claim. For example, Ayo Bamgbose (1984) reports 
on projects in Nigeria and Niger. One project (started in 1971) involved 
Hausa being used as the medium of instruction for the first three years of 
primary school. The results showed that pupils in these schools were more 
confident in their abilities; they attained literacy faster in English and 
Arabic; they were more fluent in the second language; and they acquired 
mathematical and scientific concepts faster than those in the English medium 
schools. Another project in Nigeria (1970), involves Yoruba being used as 
the medium of instruction for the first six years of primary school (a con- 
trol group had Yoruba as the medium of instruction for 3 years). The re- 
sults showed that the 6-year group performed better in math and science than 
the 3-year group and that their English language skills were equal to or 
higher than those of the 3-year group. Similar results were obtained (Bamg- 
bose 1984) from such experimental studies in the Republic of Niger where the 
mother tongues were used as media of instruction in the first three years of 
primary schools. The pupils were more fluent in French than their control 
group. In the United States (Fishman 1984), groups of native French-speak- 


ing school children in Maine showed excellent results in English when their 
initial schooling had been in French. Not only does education in mother 
tongue foster self-confidence, it has also been shown to be more effective 
in transmitting what Moumouni (1968) called, "The basic knowledge needed by 
every individual in our times." 

Finally, the Southern language policy is more practical and plausible 
than its counterpart in that it plays a very important political role. It 
is looked at by the Southerners as an element of socio-cultural and poli- 
tical identification of their African ethnicity as opposed to the assimila- 
tion policy of Arabicization. Resnick (1968:17) expresses this feeling 
vividly when he states: 

Education in the mother tongue is important 
because it is one of the chief means of pre- 
serving whatever is good in native customs, 
ideas and ideals, and thereby preserving 
what is more important than all else: namely, 
native self-respect. 

It is this "native self-respect" that the Southerners are trying to 
preserve through the development and preservation of their vernaculars. 
Achieving this right leads to a political stability of the country and, 
therefore, economic prosperity. The Addis Ababa Agreement 0972) has shown 
that for the good of the Sudanese nation peace is better than war; tolerance 
is better than repression, and the policy of reconciliation is far better 
than the policy of totalitarianism. Absolute Arabicization is a direct con- 
tradiction to the Sudanese aspirations as a nation. 

The Southern language policy was formulated by people who were fully 
aware of the importance of the role of the Arabic language as an instrument 
of national integration. That is why its status stands high in the Southern 
language policy; not because the Southerners are submitting to the total 
package of Arabicization as some people would like to see. At the same time, 
English is maintained in the system because of its important role as a link 
between the Sudanese and the rest of the African countries and the world at 
large. This is a role that is acceptable even to the Khartoum government 
which has allowed English to be taught as a subject in the schools. 


I have tried to develop the picture that the general campaign by the OAU 
member states for the promotion of the indigenous African languages to be 
adopted as media of instruction in schools should not be abused as the case 
in the Sudan shows; that the Khartoum Government in the Sudan is deviating 
from the OAU concept of indigenizat ion of the language policies. The Govern- 
ment has been leaning toward the adoption of a totalitarian language policy 
based on the great policy of Arabicization which simply means socio-cultural 
and linguistic assimilation of the indigenous African groups. This policy 
has failed in the South since it seeks to replace the highly valued vernacular 
languages. In the North, Beja, Nubian, Fur, Masalit, etc., still constitute 


50%, 19%, 2 1%, 12.5% of the population, respectively. But unlike the situa- 
tion in the South, none of these languages are used in teaching children in 
primary schools. Instead, Arabic is used indiscriminately on the assumption 
that the Northern societies have adopted Arabic as their mother tongue. 
Such a picture is false and should not be encouraged. 

All that the African countries need is a gradual evolution of national 
languages without a dramatic imposition or denunciation of the vernaculars 
of the minorities which still serve as means of communication in rural areas 
where the majority of our children go to school. These children should be 
given the chance to learn how to read and write in their mother tongues. 


Parliamentary Proceedings: Second Sitting of the First Session of Par- 
liament, 1958, p. 3. Quoted from Alibino, p. 6. (0. Alibino, The Sudan: A 
Southern Viewpoint, London, 1970). 

These people are referred to as Negroids or Southerners as opposed to 

the Northerners. 

Sudan African Closed Districts National Union (1963), Petition to the 
United Nations. 

(i) Beshir, M.O. (1968). "The Southern Sudan. Background to Conflict." 
1930 Memorandum of Southern Policy (Appendix 1:115). (ii) However, this po- 
licy was later on reversed in 1946, allowing for the unity between the South 
and the North. 

Jaden, Agrey. Khartoum Conference on Southern Sudan (March, 1965). 
Sudan Government, Kliartoum. 

Muddathir, Abd al-Rahim (1968). "Arabism, Africanism and Self-Identi- 
fication in the Sudan." Sudan in Africa, edited by Yusuf Fadl. Khartoum. 

See proceedings of the Sudan Constituent Assembly, October, 1966. 
Quoted from Wai, p. 24. (D.M. Wai, ed . , The Southern Sudan and the Problem 
of National Integration, 1973. 

'Education in the Sudan', Eleventh Annual Conference of the Philoso- 
phical Society of the Sudan (1963), p. 3, in a paper by S.V. Rao, "Adult 
Education in the Sudan." 


The Secretary, who presumably drafted the Report, was a Sudanese offi- 
cial of the Ministry of Education, Dr. Ahmad al-Tayyid. 


The Addis Ababa Agreement on the Problem of South Sudan. Draft 
organic law to organize regional self-government in the Southern provinces 
of the Democratic Republic of the Sudan, 1972. Quoted from Wai, ibid, 
Appendix VII. 

'Report of the Commission of Inquiry into the Disturbances in the 
Southern Sudan during August 1955' (Khartoum 1956:87). 

12 . . ... 

Pre-university education period in the Sudan is split into two 

stages of six years: primary and secondary (6=3+3). Secondary stage 

is subdivided into: general and academic, with three years each. Primary 

education is six years. 













0/\R run 










The Language Policy of the Southern Sudan. The High Executive Council's 
Resolution No. 273, November 1975. Southern Sudan. 

A. In the case of rural schools: 

1. The vernacular be used as medium of instruction in the first 
and second years with Arabic and English introduced orally; 

2. The vernacular be used as medium of instruction in the third 
and fourth years while Arabic and English are intensified; 

3. Arabic be the medium of instruction in fifth and sixth years 
while English continues to be intensified. 

B. In the case of urban areas: 

1. Arabic be the medium of instruction in the first and second 
years while English is introduced orally; 

2. Arabic continues as medium of instruction in third and fourth 
years while English is introduced in writing; 

3. In fifth and sixth years Arabic continues as medium of in- 
struction while English is intensified. 

C. In all junior secondary schools Arabic shall be the medium of 
instruction while English is intensified. 

D. In all senior secondary and post senior secondary institutions, 
English shall be the medium of instruction and Arabic is taught 
as a language with its literature. 

E. Adult education shall be conducted in local languages and in 
Arabic . 

F. The Regional Ministry of Education shall establish an institute 
for regional languages with a department for local languages and 
shall seek the assistance of the Summer Institute of Linguistics 
in the development of the Southern local languages. 

G. Specialization in local languages at senior secondary and higher 
institutional levels on optional bases be encouraged. 

H. The Regional Government supports the establishment of the Re- 
gional Curriculum Development Center. 

I. The decision will be implemented gradually taking into account 
the present pattern of education which allows for both English 
and Arabic as a medium of instruction; and 

J. The language problem be open for wide discussions. 



ALIBINO, E. 1970. The Sudan: A Southern Viewpoint. Oxford: Oxford Uni- 
versity Press. 

ANDREJEWSKI, B.W. 1968. The Study of the Bedauye Language: The Present 
Position and Prospects. Khartoum: Khartoum University Press. 

BAMGBOSE, A. 1984. "Mother-Tongue Medium and Scholastic Attainment in 
Nigeria." Prospects 14,1:87-93. 

BESHIR, M.O. 1968. The Southern Sudan: Background to Conflict. Oxford: 
Oxford University Press. 

BESHIR, M.O. 1969. Educational Development in the Sudan 1898-1956. Ox- 
ford: Oxford University Press. 

BOKAMBA, E. and J. Tlou. 1976. "The Consequences of the Language Poli- 
cies of African States Vis-a-Vis Education." In Reconsideration of 
African Linguistic Policies. Kampala, Uganda: OAU/BIL 1980, Publi- 
cation 3. 

COWAN, J.R. 1984. "Literacy in the Southern Sudan: A Case Study of the 
Variables Affecting Literacy Programs." In Annual Review of Applied 
Linguistics 1983, Volume 4, pp. 75-92. Rowley, MA: Newbury House 
Publishers, Inc. 

COWAN, J.R. and G.A. Cziko. 1984. Evaluation of the Southern Sudan Local 
Languages Literacy Project: Final Report. University of Illinois at 
Urbana-Champaign . 

EDUCATION IN THE SUDAN. 1963. Eleventh Annual Conference of the Philoso- 
phical Society of the Sudan. Khartoum. 

FISHMAN, J. A. 1984. "Minority Mother Tongues in Education." Prospects 

GARANG, J.U. 1973. "On Economics and Regional Autonomy." In D.M. Wai, 
ed.. The Southern Sudan: The Problem of National Integration. Lon- 

HENDERSON, K.D.D. 1965. Sudan Republic. London. 

HURREIZ, S.H. and H. Bell, eds. 1975. Directions in Sudanese Linguistics 
and Folklore. Khartoum: Khartoum University Press. 

JERNUDD, B. 1979. The Language Survey of Sudan. The First Phase: A 
Questionnaire Survey in Schools. UMEA. 

KACHRU, B.B. 1982. The Other Tongue: English Across Cultures. Urbana: 
University of Illinois Press. 

KALEMA, J. and K. Mateene, eds. 1980. Reconsideration of African Linguis- 
tic Policies. Kampala: OAU/BIL 1980, Publication 3. 

KAMBAL, M.A. 1980. An Analysis of Khartoum University Students' Composi- 
tion Errors with Implications for Remedial English in the Context of 
Arabicization. Unpublished Ph.D. dissertation. The University of 
Texas at Austin. 

MAHMUD, U.A. 1973. "The Language Situation in Deim Zubeir." Khartoum: 
University of Khartoum. Unpublished article. 

MAHMUD, U.A. 1983. Arabic in the Southern Sudan: History and Spread of 
Pidgin-Creole. Khartoum: Khartoum University Press. 

MCLOUGHLIN, P.F.M. 1962. "Economic Development and the Heritage of Sla- 
very in the Sudan Republic." Africa 32. 

MCMILLAN, M. 1970. "Aspects of Bilingualism in University Education in 
the Sudan." In T.P. Gorman, ed . , Language in Education in Eastern 


MOUMOUNI, A. 1968. Education in Africa. New York: Praeger and Sons. 

MUDDATHIR, A.A.-R. 1968. "Arabism, Africanism, and Self-Identification 
in the Sudan." In Fadl, Y., ed., Sudan in Africa. Khartoum: Khar- 
toum University Press. 

MUHMMED, A.O. 1972. Beja Texts. Khartoum: Khartoum University Press. 

I-aiRRAY, W.A. 1963. "English in the Sudan: Trends and Policies, Rela- 
tions with Arabic." In Spencer, ed.. Language in Africa. London: 
Cambridge University Press. 

NARS El Hag Ali. 1960. "Education Problems in the Sudan." Sudan Notes 
and Records 41. 

NHIAL, A.A.J. 1975. "Ki-Nubi and Juba Arabic: A Comparative Study." 
In Hurreiz and Bell, eds . , Directions in Sudanese Linguistics and 
Folklore, pp. 81-93. Khartoum: Khartoum University Press. 

1930 MEMORANDUM ON SOUTHERN POLICY. Civil Secretary's Office, Khartoum, 
January 25, 1930. In Beshir, The Southern Sudan: Background to 
Conflict, Appendix I. Oxford: Oxford University Press, 1968. 

ODUHO, J. and W. Deng. 1970. The Problem of the Southern Sudan. London: 
Oxford University Press. 


RESNICK, I.N., ed. 1968. Tanzania: Revolution by Education. Arusha: 
Longman of Tanzania. 

SANDERSON, L.P. and N. Sanderson. 1981. Education, Religion and Politics 
in Southern Sudan 1899-1964. London: Ithaca Press. 

SHINNIE, P.L. 1978. The Ancient History of the Sudan. London. 

STEVENSON, R.C. 1962. "Linguistic Research in the Nuba Mountains I, II." 
Sudan Notes and Records 43:79-102, 118-150. 

Juba. NO SG/HEC/SLR/1 .a.2. 1975. Juba, Sudan. 

THE REPUBLIC OF THE SUDAN. 1955. Report of the International Commission 
on Secondary Education. 

THELWALL, R. 1978. "Linguistic Profile." In Thelwall, R., ed.. Aspects 
of Language in the Sudan. University of Coleraine Press. 

TUCKER, A.N. 1934. "The Linguistic Situation in the Southern Sudan." 
Africa 7, 1 :31-4. 

WAI, D.M., ed. 1973. The Southern Sudan: The Problem of National Inte- 
gration. London: Cass. 

YOKWE, E.M. 1977. Polataka Linguistic Course. A Report to the Vice- 

Chancellor of the University of Juba: An Inquiry into the Polataka 
Linguistic Course. Juba, Sudan. Unpublished manuscript. 

YOKWE, E.M. 1984. "The Diversity of Juba Arabic." Paper presented to 

the 15th Annual Conference of African Linguistics at UCLA, March 29- 
31, 1984. 

The following special issue is in preparation: 

Linguistic Studies in Memory of 

Theodore M. Lightner 
Editors: Michael J. Kenstowicz 
Charles W. Kisseberth 


The following issues are available: 
Spring 1977 Vol. 7, No. 1, Papers in General Linguistics 

Fall 1977 Vol. 7, No. 2, Studies in East Asian Linguistics 

Editors: Chin-chuan Cheng and 
Chin-W. Kim 

Spring 1978 Vol. 8, No. 1, Papers in General Linguistics 

Fall 1978 Vol. 8, No. 2, Linguistics in the Seventies: 

Directions and Prospects 
Editor: Braj B. Kachru 

Spring 1979 Vol. 9, No. 1 , Papers in General Linguistics 

Fall 1979 Vol. 9, No. 2, Relational Grammar and Semantics 

Editor: Jerry L. Morgan 

Spring 1980 Vol. 10, No. 1, Papers in General Linguistics 

Fall 1980 Vol. 10, No. 2, Studies in Arabic Linguistics 

Editor: Michael J. Kenstowicz 

Spring 1981 Vol. 11, No. 1, Papers in General Linguistics 

Fall 1981 Vol. 1 1 , No. 2, Dimensions of South Asian 

Editor: Yamuna Kachru 

Spring 1982 Vol. 12, No. \, Papers in General Linguistics 

Fall 1982 Vol. 12, No. 2, Papers on Diachronic Syntax: 

Six Case Studies 
Editor: Hans Henrich Hock 

Spring 1983 Vol. 13, No. I, Papers in General Linguistics 

Fall 1983 Vol. 13, No. 2, Studies in Language Variation: 

Nonwestern Case Studies 
Editor: Braj B. Kachru 

Spring 1984 Vol. 14, No. \ , Papers in General Linguistics 

Fall 1984 Vol. 14, No. 2, Language in African Culture and Society 

Editor: Eyamba G. Bokamba 


Orders should be sent to: 

SLS Subscriptions, Department of Linguistics 

4088 Foreign Languages Building 

707 S. Mathews 

University of Illinois 

Urbana, Illinois 61801 

Studies in 

The Linguistic Sciences 









DIANA ARCHANGELI CV-skeleton or X-skeleton: the Turkish 

evidence 1 

JEAN D'SOUZA Schwa syncope and vowel nasalization in Hindi-Urdu: 
a non-linear approach 11 

ANDREA S. DUNN Swahili policy implementation in Tanzania: the role 
of the National Swahili Council (BAKITA) 31 

HANS HENRICH HOCK Yes, Virginia, syntactic reconstruction is 

possible 49 

OMAR KA Syllable structure and suffixation in Wolof 61 

YAMUNA KACHRU Applied linguistics and foreign language teaching: 
anon- Western perspective 91 


languages: implications for relational grammar 1 09 

TSUNEKO NAKAZAWA How do tense and aspect interact in 

determination of verb forms? Verb past forms and non-past 
forms in Japanese 'when'-clauses 135 

JON ORTIZ DE LIRBINA7jiJrt*||||B«s^r«Mtii unaccusativity and 

ergativity .' 147 



d nonlinear 


TAMARA VALENTINE Sex, power and linguistic strategies in the 

Hindi language 195 

Department of Linguistics 
University of Illinois 





EDITORS: Charles W. Kisseberth, Braj B. Kachru, Jerry L. Morgan 

REVIEW EDITORS: Chin-W. Kim and Ladislav Zgusta 

EDITORIAL BOARD: Eyamba G. Bokamba, Chin-chuan Cheng, Peter 
Cole, Alice Davison, Georgia M. Green, Hans Henrich Hock, Yamuna 
Kachru, Henry Kahane, Michael J. Kenstowicz and Howard Maclay. 

AIM: SLS is intended as a forum for the presentation of the latest original 
research by the faculty and especially students of the Department of 
Linguistics, University of Illinois, Urbana-Champaign. Especially invited 
papers by scholars not associated with the University of Illinois will also be in- 

SPECIAL ISSUES: Since its inception SLS has devoted one issue each year to 
restricted, specialized topics. A complete list of such special issues is given on 
the back cover. The following special issue is under preparation: Linguistic 
Studies in Memory of Theodore M. Lightner, edited by Michael J. Kenstowicz 
and Charles W. Kisseberth. 

BOOKS FOR REVIEW: Review copies of books (in duplicate) may be sent to 
the Review Editors, Studies in the Linguistic Sciences, Department of 
Linguistics, University of Illinois, Urbana, Illinois, 61801. 

SUBSCRIPTION: There will be two issues during the academic year. Requests 
for subscriptions should be addressed to SLS Subscriptions, Department of 
Linguistics, 4088 Foreign Languages Building, 707 S. Mathews, University of 
Illinois, Urbana, Illinois 61801. 

Price: $5.00 (per issue) 
Price increase to $6.00 
beginning with Vol. 16:1, 



Charles W. Kisseberth 
Braj. B. Kachru, Jerry L. Morgan 


Chin-W. Kim and Ladislav Zgusta 


Eyamba G. Bokamba, Chin-chuan Cheng, Peter Cole, AHce Davison, 

Georgia M. Green, Hans Henrich Hock, Yamuna Kachru, Henry Kahane, 

Michael J. Kenstowicz and Howard Maclay. 

SPRING, 1985 



Diana Archangeli: CV-skeleton or X-skeleton: the Turkish evidence. . 1 

Jean D'souza: Schwa syncope and vowel nasalization in Hindi-Urdu: 

a non-linear approach 11 

Andrea S. Dunn: Swahili policy implementation in Tanzania: the role 

of the National Swahili Council (BAKITA) 31 

Hans Henrich Hock: Yes, Virginia, syntactic reconstruction is possible 49 

Omar Ka: Syllable structure and suffixation in Wolof 61 

Yamuna Kachru: Applied linguistics and foreign language teaching: 

a non-Western perspective 91 

Nkonko Mudipanu Kamwangamalu: Passivization in Bantu languages: 

implications for relational grammar 109 

Tsuneko Nakazawa: How do tense and aspect interact in determination 
of verb forms? Verb past forms and non-past forms in Japanese 
\jhen' -clauses 135 

Jon Ortiz de Urbina: Partitive constructions, unaccusati vi ty and 

ergativity 147 

Hyang-Sook Sohn: Korean irregular verbs and nonlinear phonology . . . 157 

Tamara Valentine: Sex, power and linguistic strategies in the Hindi 

language 195 

studies in the Linguistic Sciences 
Volume 15, Number 1, Spring 1985 

Diana Archangeli 

In recent work on s)'llab1e structure and the core 
skeleton. Levin (1983, 1984) argues that the core skeleton is 
defined in terms of unlabeled positions or slots ("X"s) and 
not in terms of slots labeled "C" or "V". Here we address 
one of the most compelling arguments for the C and V labels, 
Clements and Keyser's (1983) discussion of Turkish "empty 
consonants". Adopting proposals found in Levin's work, I 
argue that regularities found in the Turkish data are readily 
expressed without recourse to labeling skeletal slots, and 
with rules no more complex than those reguired with the C and 
V labels. This demonstration, along with Levin's work on 
Klamath, leads to the conclusion that the set of phonological 
primitives need not include the C and V labels for skeletal 

In developing a three-dimensional model of syllable structure, 
Clements and Keyser (1983) (henceforth C&K) argue that certain nominal 
paradigms in Turkish provide evidence for labeling the core skeleton 
with Cs and Vs. I demonstrate here that the Turkish data can be 
accounted for without reference to the slot labels C, V (yet the 
essence of C&K's analysis is kept intact). This extends work by Levin 

(1983, 1984) in arguing against a labeled core skeleton. First the 
data and the CV-less analysis are presented. The theory used here and 
that advanced in C&K are then compared. Through this comparison we see 
that by limiting core syllabification to the construction of onsets 
(i.e. of "CV" syllables but not "CVC" syllables), we are able to 
dispense with the C, V labels. This same move is argued for 
independently in Steriade (1982) and Levin (1983, 1984). 

The paradigm of interest is presented in (1) below (reproduced 
from C&K: 67). In (la) the nouns are vowel-final, and in (lb) they are 
consonant-final, as seen in the nominative cases. The dative and 
possessive affixes vary, depending on the final segment in the noun: 
the vowel-initial affix -Iniz loses its initial vowel when concatenated 
with vowel-final nouns (la) and the consonant-initial affixes -yE and 

-si lose the initial consonant after consonant-final nouns (lb). 
Straightforward segmental deletion rules account for the allomorphy. 


nom. . 




























With nouns ending in long vowels, we find a contrast. Some (2a) 
pattern like the (short-)vowel-final nouns in (la) and others (2b) 
pattern with the consonant-final nouns of (lb) (reproduced from C&K: 
69, 68 respectively) : 



la la: 
(musical note) 





spelling imla: 





building bina: 



bina : si- 



mountain da: 





avalanche c4-: 





dew ci: 




c i i n i z 

The contrast between (2a), (2b) may be attributed to different 
underlying representations. C&K propose the following (C&K: 70); 


C V V 

I ^ 

1 a 

C V c 


The label of the final slot (C or V) determines which allomorphy rule 
applies. (C&K never make these rules explicit. In the discussion of 
(9) and (10), I will attempt formalization of the rules C&K assume.) 

Let us remove the C, V labels and replace them by syllable 
structure sufficient to differentiate the two representations above, 
following Levin. The "VV" of (3a) is represented as a two-part nucleus 


and the "VC" of (3b) as a nucleus followed by an unsyllabified slot: 

a. K b. I 


la da 

Associations between segments and skeleta include a rule of 
Syllable-Internal Spread (SIS, 5 below) as well as the Universal 
Association Convention (UAC). I follow Pulleyblank (1983) in assuming 
that multiple linkings dre not an automatic consequence of the UAC. 
Consequently, a rule, SIS, is needed to spread the vowel matrix to the 
second nuclear position. 

(5) Syllable Internal Spread (SIS) 


Automatic Core Syllabification creates onsets on all syllables, if 

(6) Core Syllabification (CS) 


CS (6) is part of Universal Grammar, of the same theoretical status as 
the UAC. This captures the observation that all languages have 
syllables with onsets, although not every language has syllables with 
codas and onsetless syllables do exist. Also, like the UAC, it applies 
whenever possible, thus accounting for resyllabification into onsets. 

The forms from (4) are given a CV-less representation in (7), and 
corresponding short vowel-final and consonant-final nouns are 
represented in (6). In (7), (8) the UAC, SIS (5), and CS (6) have 
taken effect. Results of the UAC and CS are given by solid lines; 
results of SIS (5) are represented by the dashed line. 

A '■ A 


la da 


I A '-A 



d a sap 

Examination of the representations in (7) and (8) reveals that the 
environments of the allomorphy rules are expressible in terms of 
syllable structure. The morpheme-initial segment deletes after a 
syllabified slot with the affix -Iniz : 

X — * / X_] where _X is a syllabified slot 


and it deletes after an unsyllabified slot with the affixes ^^ and 
-si : 

X — > / X'] where X' is an unsyllabified slot 


Once these allomSrphic deletions take place, syllabification proceeds 
according to the following rule: 

(11) Coda Rule: 


(Rules/conventions similar to CS (6) and CR (11) are found in C&K, 
Harris (1983), Steriade (1982), to name a few.) As noted above, I 
follow Steriade (1982), Levin (1983, 1984) in ascribing a universal 
status to CS (6) and a language-particular status to CR (11). 

The following derivations produce results that are essentially 

identical to those in C&K, except for the lack of CV labels on the core 

skeleton. In the first step, the UAC provides for one-to-one, 

left-to-riqht association of the melody to the skeletal slots and CS 
(6) builds onsets. These are both shown by dashed lines in the first 
step of (12). 


UR (4), UAC, SIS (5), CS (6) 

A A A A 


II I I / 11 11/ 

• I ' '/' ' ' ' ' /' 

da la da la 

Affixation (and CS 6) 

/ /1 A A / A A A I ,■■1 


II II 1 1/ I I II ! I I I I 1/ I I I I 

da ya la ya da ♦n^-z la vn;.z 

Affixes are added and again CS (6) builds onsets wherever possible 

(shown with dashed lines). In the case of the vowel-initial affix 

-Iniz , an onset is formed from the unsyllabif ied final X of the root 

daX. In the third step, the rules of allomorphic deletion (9) and (10) 
remove the affix-initial slot in daX+yA and laX+Inlz . CS (6) applies 
wherever it can. 

(12) cont. 

allomorphic deletion (and CS 6) 

A A A / 

X X X] X n/a n/a X X X] X X X 

III I 1/ I M 

d a a 1 a n + z 

CR (11) 

AAA: A A: 

n/a n/a X X X] X X X X X X X] X X X 

II I I I I 11/111 

da i-n + z la ni-z 

CR (11) now incorporates the final consonant of -Iniz into the 
preceding syllable, and the surface representations of all four forms 
are derived. 

(12) cont. 

surface representation 

// A/ AAA AA 


III 11/11 II 1 1 1 1 11/111 

da a laya da*n*z lan*z 

With daa , dainiz, a slot is syllabified as an onset even though it has 
no features associated with it. This slot, therefore, is not 
pronounced. Below, we see derivations where the empty slot is 
syllabified in the first syllable, not the second: 


UR (4) and affixation, UAC, CS (6), SIS (5) 

A I / 


da d a 1 a r 

CR (11) 

A A. A-. 


II 1 1 1 1 1 

da d a 1 a r 

SIS (5) 

A A A 


d 5 d a 1 a r 

The long vowel is derived via CR (11) and SIS (5). Thus, we may 
conclude that the CV-less theory accounts for the Turkish data. Let us 
now compare the CV-theory of C&K with the CV-less theory, to see how 
they differ, beyond the use of labels on the core skeleton. 

Both theories include syllable structure in underlying 
representation. As we have seen, underlying representations are formed 
in terms of X's with or without syllable structure in the CV-less 
theory. In C&K, syllable structure is supplied to lexical 
representations in accordance with a construction algorithm which 
states "V-elements dre prel inked to d's" (C&K: 38). Thus, underlying 
representations consist of strings of C, V which automatically become 
strings of 

C, V 

Since any V in underlying representation is linked to a syllable by 
this algorithm, anywhere an underlying V is found, we have, in essence, 
underlying syllable structure. (The V label is still essential in C&K 
as there is no internal syllable structure available to distinguish 

heads from non-heads without the labels.) 

The theories reguire the same operations. Both models include 
rules of syllabification which create onsets and codas, and include 
rules of resyllabif ication, syllable-internal spread, and allomorphy 
for Turkish. 

The Turkish al lomorphy rules are essentially eguivalent. Although 
the rules are not made explicit in C&K, we may assume something like 
the following corresponding to our (9) and (10): 


a. V— ♦0/V b. C— »0/C 

both in the appropriate morphological environments 

It is not immediately obvious that these rules are any more complex or 
any more simple than the rules in (9) and (10) — on this score also, 
the two accounts are basically the same. 

The point of difference is in the definition of core 
syllabification. In the CV-less theory, core syllabification forms 
"CV" syllables only. Syllables are closed by language particular coda 
rules. In the CV-theory, core syllabification includes coda 
formation. However, this rule is in no direct way related to the 
labels C, V. It simply characterizes a syllabification process. With 
respect to the Turkish data, we may order the coda rule with core 
syllabification or in the phonology. However, if the coda rule is part 
of core syllabification, the labels C, V are needed to determine which 
allomorphy rule is applicable. If it is not obligatorily part of core 
syllabification, then the labels C, V are unnecessary, and allomorphy 
is determined by the syllabification status of the noun-final slot. 

There are certain theoretical consequences obtained by eliminating 
the labels C, V. Most importantly, we decrease the number of primitives 
in the set of phonological symbols. Having fewer elements to 
manipulate results in fewer permutations of those elements and 
consequently severely restricts the set of possible grammars. On the 
other hand, the coda rule is a manipulation of the phonological 
primitives. Altering its role in Universal Grammar has ramifications, 
as seen here, but does not lead clearly to a reduction or an increase 
in the number of possible grammars. 

In the model proposed here, core syllabification comprises only 
what is truly universal, the creation of "CV" syllables (by CS (6)). 
This, along with the Universal Association Convention, is assumed to 
apply whenever possible, from whence we derive resyllabification into 
onsets. The coda rule (11) is language particular, as not all 
languages have coda rules, and is not part of core syllabification. 

Finally, the redundancy in underlying representations of 
prelinkinq all V slots to syllables does not exist. Certain slots are 
designated as part of a syllable, others are not. Those which are not 
syllabified in underlying representation may be incorporated in a 
syllable by core syllabification or by other syllabification rules. 


C&K advance three arguments for C, V labels, from Turkish, 
Finnish, and Klamath. Levin (1984) deals with the Klamath evidence. 
The Finnish example succombs to a straightforward transliteration into 
a CV-less representation. This article is concerned only with the 
Turkish example. 


The capital letters refer to the high {_!_) and mid (A) vowels 
which undergo harmony. 

C&K propose a syllable plane and a nucleus plane, the latter 
formally distinct from but dependent on the former. Each has a flat 
structure, that is has no internal hierarchy, (a) below. The structure 
employed here, (b) below, has one degree of hierarchy, and no labels on 
the nodes (the structure is sufficient to determine relative positions 
within the syllable). 


X X X (X) X X X (X) 

In (b), the vertical line (1) denotes the head of the syllable and the 
angled lines (/ and \) denote periphery elements. The lower node 
corresponds to C&K ' s nuclei. The difference is that in (b), both 
syllable and nucleus have heads while in (a), we may say that the 
nucleus is the head of the syllable, but the nucleus itself has no 
head. Analyses such as Harris's (1983) treatment of Spanish diphthongs 
suggest that the head/non-head positions in the nucleus need to be 
distinguished. I am grateful to Jean d'Souza for helpful discussion of 
this point. 

Nothing is said about VV seguences in C&K, whether they are in a 
single syllable or in separate syllables. 



CLEMENTS, G.N. and S.J. Keyser. 1983. CV phonology: a generative 

theory of the syllable. Linguistic Inquiry Monograph 9, 

Cambridge, MA: The MIT Press. 
HARRIS, J.W. 1983. Syllable structure and stress in Spanish: a 

nonlinear analysis. Linguistic Inguiry Monograph 8, Cambridge, 

MA: The MIT Press. 
LEVIN, J. 1983. 'Reduplication and prosodic structure', ms. 
LEVIN, J. 1984. 'Conditions on syllable structure and categories in 

Klamath phonology'. Proceedings of WCCFL 1984. 
PULLEYBLANK, D. 1983. Tone in lexical phonology. Cambridge, MA: 

doctoral dissertation, MIT; to be published by Dordrecht, Holland: 

Reidel Press. 
STERIADE, D. 1982. Greek prosodies and the nature of syllabification. 

Cambridge, MA: unpublished doctoral dissertation, MIT. 

studies in the Linguistic Sciences 
Volume 15, Number 1, Spring 1985 

Jean D'souza 

Schwa syncope in Hindi-Urdu has been of interest to phono- 
logists for some time and several linear analyses have been 
proposed to account for it. The phenomenon is made more 
interesting because of its interaction with vowel nasalization 
which suggests that the nasal vowels in Hindi-Urdu should be 
treated as a sequence of long vowel and nasal consonant 
rather than as an underlying nasal vowel. Provided such an 
analysis is possible, a considerable economy in the phonemic 
inventory of the language can be effected. In the linear analyses 
of schwa syncope, what started out as a simple rule: a --) 0/VC 
(Pray 1970) , became more and more complex in order to deal with 
all the data. Recent developments in non-linear phonology have 
made possible a simpler and more insightful account of the Hindi- 
Urdu facts. Chief among these are: 1) the development of the 
core skeleton as a series of unlabelled slots rather than a series 
of CV elements. 2) the use of the foot as a domain for segmental 
processes. In the present analysis, syllable structure and 
syncope feet make possible an account which is more elegant and 
more illuminating than any linear account. 

1.0 Introduction. 

Schwa syncope is a very productive rule in Hindi-Urdu and has been 
the sxobject of considerable interest for a long time. Several different 
accounts of schwa deletion have been proposed, e.g.. Pray (1970), 
Narang and Becker (1971) , Ohala (1972) , among others. All these analyses 
have dealt with schwa syncope in linear terms. This paper presents a non- 
linear analysis of schwa deletion and shows that a non-linear analysis 
in terms of syllable and foot structure provides a simple and insightful 
account of the phenomenon of schwa deletion. Some of the other accounts 
proposed for schwa syncope are exeunined and it is shown that a syllable 
based analysis does not encounter any of the difficulties the linear analyses 
run into. More importantly, facts which have to be accounted for by 
stipulations in a linear analysis or demand the use of global powers 
are shown to fall out very naturally from syllable structure. 

Finally, vowel nasalization, a much debated issue in Hindi-Urdu 
phonology, is looked at and it is suggested that a non-linear analysis 
not only sheds light on the interaction between schwa deletion and 
vowel nasalization, it also makes possible an account that effects a 
considerable economy in the phonemic inventory of the language. 

1.1 The Facts. 

Schwa deletion is a very general rule in Hindi-Urdu. It deletes 
the final schwa of a verb stem if that stem is followed by a vowel 
initial suffix. For example: 



(l. ) verb stem past gloss 

n i ka I n i k I a : ' come out ' 

pakar pakra: 'catch' 

pi^ak picka: 'squeeze' 

The rule also applies to noun stems. For example: 

(2.) nom. sing oblique pi. gloss 

kamar kamro: 'waist' 

sabak sabko: 'lesson' 

ke: sar ketsro: 'saffron' 

Evidence for this being a rule of deletion (vs. insertion) comes from 
forms in the nominative singular which end in consonant clusters. For 
example in the forms given below no schwa intervenes between the under- 
lined clusters. 

(3.) qat I qatlo: 'murder' 

swarg swargo: 'heaven' 

kalt ka?to: 'trouble' 

Syncope does not apply if the schwa is preceded by a consonant 

(4.) pustak pustako: 'book' 

ki:rtan khrtano: 'song' 

aksar aksaro: 'letter' 

In addition it does not apply if the schwa is followed by a consonant 

(5.) palaQg palaogo: 'bed' 

tilasm tilasmo: 'magic' 

daraxt daraxto: 'tree' 

1.2 Previous Analyses. 

These facts led Pray (1970) to formulate the schwa deletion rule 
as follows:"* 

(6.) a — > 0/VC CV 

Narang and Becker (1971) elaborated this rule a bit in order to 

stop schwa deletion from applying word internally. They did this to 

account for the word [wa:ra:nasi : ] the old form of the name Benaras . 
Their rule is:^ 





-■> 0/VC C + V 

This inclusion of the morpheme boundary to the right is not a 
crucial addition as [wa:ra:nasi : ] seems to be the only word that demands 
it and hence can be treated as an exception. In addition, as pointed 
out by Ohala (1972) schwa syncope does optionally apply morpheme 


internally in some cases and the presence of the morpheme boundary 
renders Narang and Becker's rule incapable of accounting for alterna- 
tions like the following: 

(8.) rajan i : '^ rajni: 'night' 

kaidambari i^ ka:dambri: 'a novel' 

To account for these alternations and also for the fact that schwa 
deletion optionally applies in the speech of some native speakers in 
spite of a preceding consonant cluster, Ohala (1972) elaborates the rule 
still further. She also takes into consideration the sociolinguistic 
contexts in which the rule applies. Her rule is formulated as follows: 

(9.) 9 — :> 0/VcJ ^cv/)[^ 

loan / 

•i L+casual speechj y 
[ C+normal tempo 1 ( 

Condition 1: There may be no + in the environment to the left. 
Condition 2: The output of the rule will not violate the sequential 

constraints of Hindi. 
Condition 3: The rule applies from right to left. 

Ohala cites forms like j^Ogsli' j^Ogli: 'wild' and do: Qgari : 
do: ogri: 'pertaining to hill people' to show that schwa deletion does 
not apply for some speakers in spite of a preceding consonant cluster. 
She invokes Condition 1 in order to account for: a) the non-application 
of the rule in stems to which prefixes have been added, for example: 



+ par** + a: 

— * 

bepar a: 

' unread' 

*bepr i 


+ samay 

—■ ) 


' inopportune; 



+ mar +an 



'until death' 


b) its non-application in suffixes containing schwa: 

(11.) kala: + wat + i: ■) kala:wati: 'name for a girl' 

*kala:wti : 
ka:ri: + gar + i: ■> ka:ri:gari: 'craftsmanship' 

*ka: ri :gri : 
eka:ki: + pan + a: ^ eka:ki:pana: ' lonesomeness; 

*ekaki :pna: 

Condition 2 is needed in order to account for exceptions to her rule, 
i.e., cases in which a schwa does not delete when preceded by a consonant 
cluster even though the rule would predict deletion. (Ohala provides 
a list of the sequential constraints that mediate against schwa deletion.) 

(12.) kudrat 'nature' kudrati: 'natural' *kudrti: 
karwat 'side' karwate 'sides' *karwte 
citwan 'glance' citwane 'glances' *citwne 

Condition 3 is the most tentative and Ohala cites as evidence the way 
in which native speakers render unfamiliar words written in Devanagari 


(the Hindi script). For example, the word for 'adopted' is [godn8si:n] 
in Hindi-Urdu. However its written form in Devanagari is "godaneSi :n" . 
Speakers unfamiliar with the word pronounce it with the second schwa 
deleted, i.e. [godansiin]. This according to Ohala suggests that the rule 
of schwa deletion may be applying from right to left. 

Ohala 's rule is rather complex and depends on several seemingly 
arbitrary conditions without providing a satisfactory explanation as to why 
these conditions are necessary. This however is not so much a criticism 
of Ohala 's analysis as it is a criticism of the theory upon which the 
analysis is based. Linear phonology just does not provide the means 
necessary for a simpler and more satisfying account of the processes 
described above. 

2.0 A Non- Linear Approach. 

I show here that a non- linear analysis in terms of syllable structure 
provides a straightforward account of the schwa deletion facts, eliminates 
the need for conditions, simplifies the rule and at the scime time sheds 
light on the reasons for the application of the rule. In my analysis I 
assume the theory of autosegmental phonology first proposed by Goldsmith 
(1976) to account for tonal phenomena and later extended by McCarthy 
(1981) to the analysis of the non-concatenative morphology of the 
Semitic languages. 

2.1 The Background. 

Following McCarthy (1981) , I assume a core skeleton and a melody 
tier in underlying representation. However, adopting proposals made 
by Levin (1984) and Archangeli (1984) , I make use of a core skeleton 
consisting of unlabelled X slots rather than the CV slots McCarthy 
proposed. (An X skeleton has certain advantages over a CV skeleton. 
Some of these advantages will become obvious from this analysis, for 
others see Levin (1984) and Archangeli (1984).) 

In the underlying representation (UR) of a lexical item one finds 
a series of unlinked melodies and an X skeleton. The melodies are 
linked to the skeleton slots by the Universal Association Convention 
i.e. one to one, left to right. Linking obeys the Well-Formedness 
Condition which rules out crossed association lines. 

2.2 Syllable Structure. 

Syllable structure is largely rule governed. In UR syllable heads 
are denoted by a vertical line linked to an X slot. All other elements 
are unlinked. In Hindi-Urdu only vowels may link to a slot identified 
as a syllable head. Vowels are marked with the feature +N, consonants 
with the feature -N (Levin 1984). For example: 

(13.) I 

Syllable heads in Hindi-Urdu may have either of the following structures: 


(14. A.) I (B.) 

(14. A.) denotes a short vowel head, (14. B.) denotes a long vowel, the 
first element of which is the head. A rule (which probably is universal) 
spreacfe [ ] in the configuration in (14.B.). Solid lines indicate 

elements pre-specif ied in UR. All other constituents of the syllable are 
incorporated by rule and are represented by broken lines e.g., 

,...., ,^, ,B., __^^^ 

XXX xxxx 

I am suggesting a differentiation between solid lines and broken lines 

as more than a notational convenience. As the data in section 3 show, 

the distinction it makes possible between elements pre-specified in 
UR and those incorporated into the syllable by rule is crucial. 

Core syllabification is obtained by the application of universal 
rules : 

(16.) i) The onset rule which incorporates a single unsyllabif ied 
slot from the left e.g.. 

ii) The coda rule which incorporates a single unsyllabif ied slot 
from the right e.g.. 

XX ---> XX 

The onset rule applies first accounting for the universally preferred 
CV syllable structure. Thus, given the structure in (17. A.) application 
of the 'onset first' principle (Clements and Keyser 1983) results in 
(17. B.) rather than (17.C.). (I follow Steriade (1982) in modifying 
the principle as formulated by Clements and Keyser so that Universal 
Grammar allows for the incorporation of only one C by the onset first 
principle. All other C's are incorporated by language specific rules. 
The onset rule is followed by the code rule which also incorporates only 
one C. ) 

xxxx xxxx xxxx 

2.3 Foot Construction. 

The concept of the foot in linguistic description, long neglected, 
has of late come into its own. Several recent studies make use of the 


foot in segmental rules, among them being Withgott (1982), Rappaport (1984^ 
Archangeli (1984) , and Hammond (1985) . 

I suggest that in Hindi-Urdu syllabification is followed by foot 
construction. The language has syncope feet which play an important part 
in certain segmental processes. These feet are constructed as follows: 

(18.) foot rule Build left dominant quantity sensitive binary 
feet beginning at the left edge of the word. 

Being quantity sensitive the rule allows feet to have weak branches 
only in case the syllable on which the weak branch is constructed has 
a light or non-branching rime. For example, 

xxxxxxxx kxxkxxx 

1 1 { I 1 1 > / ■ ' I : \ \ ■ 

bistara 'bedding' ka nta 'wife' 

t/ T r t 

A weak branch of a foot can be constructed on the second syllable of 
the word bistara: because this syllable has a non-branching rime. The 
same is not possible for the word ka:nta: because the second syllable 
has a branching rime. 

2.4 Schwa Syncope. 

The groundwork being laid, the process that leads to schwa syncope 
can be formulated as follows : 

(20.) schwa desyllabification | 

X — > X / --- 

i i V 

The rule says that if a schwa is located on the weak branch of a 
foot the association line marking it as a syllable head is severed. 
I adopt the general convention that any element not in a syllable at 
surface structure is not phonetically realized. Therefore, unless the 
desyllabified schwa is reinstated as a syllable it does not show up on 
the surface. 

Schwa desyllabification accounts for schwa deletion in words like 
sabgk + 3:, nikal + a: , ke:sar + 5 : , given in (1) and (2) above. 
For example: 


XXJJXXXX by schwa desyllabification (20) ^XXXXXX 

sabako salialco 


Core syllabification (16) iiiunediately follows desyllabif ication and 
the stranded consonant is linked to a syllable: 

(22.) _,t. /1\ 

1 I I I ' ' ' 


It also accounts for the non-deletion of schwa in words like tilgsmS: , 
pelaOgo: , etc. (5) in which the schwa is followed by a consonant cluster 

(23.) 4,'tA '1/t'A 


I 1 I I I . 1/ I I ■ •( I • / 

':;;!!■' '1 • •: ; ;' 

tilasmo palaogo 

f r I fir 

Since feet are quantity sensitive the second syllable in these 
words cannot be on a weak branch. These structures therefore do not 
present the right environment for the application of schwa desyllabi- 
fication, the schwa remains linked to the syllable and is realized 
on the surface. 

Evidence for foot construction having to be quantity sensitive comes 
from words like those in (23) for which the wrong results are obtained 
if feet are not quantity sensitive: 


I ' I I ' I ' I I ' 

I [ 1 I 1 I 'I !■•[!'.' 

tilasmo tilasmo 

HI y \ 

PR tilasmo: *tilsmo: 

In (24. A.) quantity senstive feet have been constructed and as the 
schwa is not in the scope of a weak foot it cannot be desyllabif led 
and the right result is obtained. The feet in (24. B.) however are 
quantity insensitive and as the schwa is on a weak branch, rule (20) 
can apply giving the wrong results. 

2.5 Dialectal Variation. 

Words like pustako: (4) lead us to consider the fact (mentioned 
earlier) that though for many speakers schwa does not delete when 
preceded by a consonant cluster, for an increasing number of speakers 
schwa deletion is beginning to apply even in the environment of a 
preceding cluster. Let us call the first case Dialect A and the second 
Dialect B. We will first consider Dialect A and the non-deletion 
of schwa in words like pustak + o: , ki : rtan -i- o: , etc. (4) . 


xxxxxxxx xxkxxx)ix 

!•!!:•!/ ■':;:!■'' 

pustako Icirtenc? 

\/\ v\ 

These forms provide the environment for schwa desyllabif ication. It 
applies giving: 


xxxxxxxx XxkxxxM 


I \l , , 

I I I I I i u I I J 1 1 [ 1/ 

pustslto kirtano 

V \ y I 

Now the core syllabification rules apply. First the onset rule tries 
to incorporate the stranded consonant into the syllable following it. 
This syllable however already has an onset so nothing happens. The 
coda rule also fails since the preceding syllable already has a coda. 
I suggest that at this point another rule applies reinstating the 
schwa as a syllable head. This rule can be formulated as follows: 


(27.) schwa reinstatement X ^> X / X' 

I I 

9 3 

(x' denotes an unsyllabif ied slot) 
Application of schwa reinstatement to the forms above gives: 

I 1 1 1 1 I w [III,!' 

pustalco kirtano 

Core syllabification now applies giving: 

xXxxXxix xxxxx^xX 

I ; I I I j 1/ t I ! 1 1 1 !' 

pustako kirtano 

PR pustako: kirtano: 

Dialect A therefore can be accounted for very easily. How are 
we to account for Dialect B in which schwa deletes in spite of a 
preceding cluster, e.g., japgli: , c uc undri : 'a kind of rat'? These 
cases can be looked at in terms of a reordering of rules. In Hindi-Urdu 


in addition to the core syllabification rules there are rules of extended 
syllabification which take additional consonants into the onset or the 
coda. These rules can be formulated as follows: 

(30. ) rules of extended syllabification 

a) X'XX(X) > XXX(X) 

b) X(X)XX' -^ X(X)X> 


These rules are necessary to account for words like: 
(31.) . ^ 

-'tV'^ if':- 


1 1 1/ I " ',''.'• 

kwa ra 'bachelor' lutf 'pleasure' 

in which the onset in one case and the coda in the other contain more 
than one element. In Dialect A rules of extended syllabification 
apply after the application of schwa reinstatement; in Dialect B 
however a reordering of these rules has taken place and extended 
syllabification rules apply before schwa reinstatement. The difference 
between the two dialects therefore can be shown as follows: 


1. core syllabification core syllabification 

2. foot construction foot construction 

3. schwa desyllabif ication schwa desyllabif ication 

4. core syllabification core syllabification 
r-5. schwa reinstatement rextended syllabification 

6. core syllabification Lschwa reinstatement 
1—7. extended syllabification core syllabification 

It may be noted that core syllabification follows every rule that 
alters syllable structure. Therefore foot construction which does not 
change syllable structure, and extended syllabification which leaves 
nothing new for core syllabification to do, are not followed by core 

The derivation of a word from Dialect A has already been given 
above. Let us now look at the derivation of a word from Dialect B. 


(33.) by core syllabification (16) 

I I > i I i i' 
and foot construction (18): Jsogsli 

t/ 1 

schwa desyllabification (20) • 5cxixxxxx 

t I • 


core syllabification (16) : 

l^ T 


extended syllabification (30) 

schwa reinstatement (27) : 
core syllabification (16) : 



Thus the same rules, with only a simple reordering, can account 
for both dialects of Hindi-Urdu. A small complication nonetheless 
does arise in the case of speakers of Dialect B who, though they delete 
schwa after preceding clusters do not do so in the case of all 
clusters e.g. , 

(34. ) matlabi: 

kudreti : 

' selfish' 
' natural' 

*m8tlbi : 
*kudrti : 

This, as mentioned earlier, is what led Ohala (1972) to impose Condition 
2 on her schwa deletion rule. 

Under the present analysis such forms are not too difficult to 
account for. As we have seen schwa desyllabif ication in Dialect B is 
followed by core syllabification and then by extended syllabification. 
Core syllabification cannot apply if the preceding coda and following 
onset positions have already been filled. In addition, in cases like 
the above, extended syllabification cannot apply either for reasons 
particular to Hindi-Urdu. Restrictions on permissable onset and coda 
clusters are extremely stringent in this language and not all consonants 
left stranded by desyllabification can be incorporated into a syllable (for 
allowed onsets and codas see D'souza 1983). In such cases schwa reinstate- 
ment which follows extended syllabification reinstates the schwa as the head 
of the syllable. Such examples provide independent support for schwa rein- 
statement. /|» -1 'K 


(35.) syllabification (16) and 

foot construction (18) 

schwa desyllabification (20) 

core syllabification (16) : 
extended syllabification (30) 

schwa reinstatement (27) : 

core syllabification (16) 


I I I I < I • ^- 


V T 


• 1 \ I 1 1 I / 


U^ T 


; ' ; ' ! I [/ 


i I I I I r I, 


Thus what a linear approach has to formulate as a condition on a 
rule is a natural result of syllabification. 

2.6 Initial Accenting. 

We have seen so far that looking at schwa deletion in terms of 
syllable and foot structure accounts for most of the Hindi-Urdu data. 
Let us now briefly examine the non- application of schwa syncope across a 
morpheme boundary to the left (Ohala's Condition 1). As noted by Ohala, 
schwa syncope does not apply in cases like the following: 

(36.) be + par + a: 'unread' 

kaia: + wati: 'name for a girl' 
a + mar + an 'until death' 

These forms can be accounted for if we assume that all morpheme 
initial syllables are accented i.e., they are heads of feet. 

(37.) initial accent Every morpheme initial syllable is 
the head of a foot. 

This assumption enables us to account for all the data in a fairly 
straightforward manner. The fact that initial morphemes are accented 
is not peculiar to Hindi-Urdu as several other languages employ the 
same strategy. (For an account of initial accenting in Tunica see 
Hammond 1984) . Given the forms above, initial accenting results in 
the following feet being built: 

(38.) be + paf + a: ka:ri: + gar + i: 

111 1 1 T I 

Schwa in these words therefore is never in the right environment for the 
application of desyllabif ication and the question of deleting it does 
not arrive. 

(Morpheme boundaries block the application of phonological rules in 
several other languages. In Arabic there is a rule of vowel deletion: 

V -■> 0/[CVC ^CVC] , which applies when there is a morpheme boundary 

to the left but not when there is one to the right (Hudson n.d.). Tonkawa 
(Kenstowicz and Kisseberth 1979) also has a rule of vowel deletion: 

V --> 0/#CVC CV which is blocked by a morpheme boundary to the right. 
Hindi therefore seems to fit a pattern and it is likely that syllable 
accenting can account for Arabic and Tonkawa as it does for Hindi.) 

2.7 Summing Up. 

So far I have shown that a non- linear analysis allows for a 
simple and insightful account of the schwa deletion phenomenon in that 
all the data are accounted for by two fairly straightforward rules, 
schwa desyllabif ication (20) and schwa reinstatement (27) , without the 
need for conditions and stipulations. The same rules with a simple re- 
ordering (32) account for dialectal differences, and with the addition 
of initial accenting (37) , for non-application of desyllabif ication 
across morpheme boundaries to the left of the schwa. 


In addition, this approach to schwa deletion suggests a possible 
account of word internal non-deletion in slow speech i.e., alternations 
like: ka : dambar i : ->^ ka : d anbr i : . Since schwa is desyllabif ied, not deleted, 
the X slot (which is a timing unit) and the schwa are present right up to 
surface structure. I suggest that in slow speech the timing slots are 
given equal importance (Hindi-Urdu is a syllable timed language) and in 
this case the X structure is reinstated as a syllable head. Of course 

since the X structure is not a syllable it will not get the same weight a 

syllable would. It does however get some weight and this triggers a 
surface level equivalent of schwa reinstatement. 

3.0 Vowel Nasalization. 

In this section I will examine the interaction of schwa deletion 
and vowel nasalization and will show how this interaction provides evidence 
for treating nasal vowels as an underlying sequence of vowel and nasal 
consonant thus reducing the phonemic inventory of the language significantly. 

The nasalized vowels in Hindi-Urdu have received varying treatments. 
Some writers claim that the language has a set of nasalized vowels in 
addition to the oral ones (Ohala 1972) while others claim that all 
nasalized vowels can be derived from oral ones (Narang and Becker 1971) . 
Narang and Becker propose that nasalized vowels be treated as an underlying 
sequence of oral vowel and nasal consonant. They advance this analysis on 
several grounds, the strongest to my mind being the fact that in Hindi- 
Urdu most nasalized vowels are long. There are very few short nasalized 
vowels in the language. The advantage of an analysis like Narang and Becker's 
over one like Ohala 's is that it allows for a considerable economy in the 
phonemic inventory of Hindi-Urdu and if viewed autosegmentally can 
account for the length of nasal vowels. 

Narang and Becker set up the following rule to account for 
nasalization in Hindi-Urdu: 

(39. ) r+syllabic| f+nasall C --- > r+nasalj 
l_+tense J 

3 12 

The interaction of schwa syncope and vowel nasalization provides 
evidence in favor of Narang and Becker's approach because schwa syncope 
seems to treat nasalized vowels as a sequence of vowel and nasal 
i.e., when a schwa is preceded by a nasalized vowel it does not delete 
just as it does not delete when preceded by a consonant cluster. 


) a:g3n 

, h~ V , ~ 
b a : j ako : 

' courtyard' 
' rope maker ' 



a: calo: 


*a: c:o: 

Narang and Becker order their rule of schwa deletion after their 
rule of nasalization. They do this so that they can account for forms 
like a;gan,a:gano: ' courtyard/s' . 



) UR 

schwa deletion 



+ 0:N, 







If nasalization applied first schwa deletion could also apply giving the 
wrong output: * a:gno: . 

However, as Bhatia and Kenstowicz (1972) point out, this rule 
ordering results in an ordering paradox. Narang and Becker cannot 
account for forms like the following if they allow schwa deletion to 
precede nasalization: 

(42.) ina:n8s matnsi: 'mind' 
ki:mat ki:mti: 'price' 
da : naw da : nwo : ' demon ' 

Words like these surface with a long oral vowel followed by a nasal 
consonant yet schva deletion has applied. Given Narang and Becker's 
rule ordering the wrong surface representations will be obtained for 
these words: 

(43.) UR /da: new + o:N/ 

schwa deletion /da:nwo:N/ 

nasalization /da:wo:/ 
PR *da:w6: 

The corect forms could be obtained if schwa deletion was ordered 
after nasalization but then the wrong output would be obtained for words 
like a:gano: . 

Bhatia and Kenstowicz (1972) suggest that the problem is one of 
derivational history, nasalization applying only to underlying V:NC 
sequences not derived ones, and advocate the use of global rules as a 
possible solution. Global rules allow one to describe derivational 
history without recourse to rule ordering. However, global rules are 
undesirable in that they add too much power to the grammar (Kiparsky 
1973) . One of the great advantages of an approach involving syllable 
structure is that the need for global rules is largely eliminated. Levin 
(1984) has shown this for Klamath, Archangeli (1984) has shown it for 
Yawelmani and this study shows it for Hindi-Urdu. The question as to what 
it is that syllable structure provides that makes global rules unnecessary 
is a question worth investigating. One possible answer comes from the 
fact that global rules are necessary in cases in which one has to refer 
to two levels of structure or in other words to two points in a 
derivation. Non- linear phonology provides this information, not as a result 
of global power, but as a consequence of the representation. As the 
evidence below will demonstrate, the fact that one can refer to several 
tiers of information at the same time is crucial to the analysis of vowel 
nasalization. Linear phonology could only get this information by adding 
power to the grammar. Autosegmental phonology gets it as a basic part of 
the theory. 


Further, problems are inherent in the way Narang and Becker 
formulate their rule. Theirs is a morpheme based rule and they crucially 
rely on it applying in the environment of a consonant to the right 
of the nasalized vowel. They therefore have to resort to ad hoc rules 
in order to account for forms like the following in which a nasalized 
vowel is found at the end of the word. 

(44. ) ma: 'mother' d ua: 'smoke' 
lerko: 'boys' ha: 'yes' 

The main defect in Narang and Becker's analysis is that they base 
their rule upon the observation that "within a morpheme long vowels 
do not occur before a nasal consonant cluster". They cite the word 
sa:nt as an exception. However, as Bhatia and Kenstowicz have pointed 
out there are several other such "exceptions" e.g., 

(45.) pra:nt 'district' 
ka:nta: 'wife' 
a:ntrik 'internal' 

Looking at nasalization in terms of syllable structure allows for 
the formulation of a rule which accounts for all the facts without falling 
foul of the ordering paradox and without having to invoke the power 
of global rules. Further, if we take the nasal vowels as being 
underlyingly short rather than long we can account for the data and also 
for the fact that there are very few short nasalized vowels in Hindi-Urdu. 

I suggest that a word like a:gan has the following underlying 
representation : 

(46. ) 


The nasal consonant is in the coda and linked to an X slot in under- 
lying representation. 

To deal with nasalization I adopt a coplanar representation of 
features (Archangeli 1984) and suggest that the feature [+nasal] is on 
the same plane but off on a separate tier from the other features: 

(47.) n 
[ ] 

I , 


A rule of nasal spread causes this feature to spread onto the preceding 
vowel in case there is a nasal in the coda in underlying representation. 


(48.) nasal spread 




The feature [+nasal] spreads to the vowel and the vowel spreads to the 
X slot in the coda by a rule rather like the (universal) rule of long 
vowel formation except that it is ordered to apply after nasal spread. 

(49.) vowel spread 


— ^ 




[ ] 


Pulleyblaak (1983) argues convincingly for the position that if 
something is linked by rule to an already linked slot the original 
association line is automatically broken. Thus vowel spread results in 
the delinking of the other features attached to X. Let us look at 
a sample derivation. (For convenience, I leave out at each stage 
information not directly relevant to that stage.) 

N ! 

(50.)UR XXXXX 

by syllabification (16) 
and foot const. (18) : 

by nasal spread 



by vowel spread (49) 



I \ 


: I ; ; : 

a n g d n 


a n g a n 

I ][ ] 





It may be recalled that the interaction of schwa deletion and 
nasalization led to an ordering paradox for Narang and Becker. Given the 
present analysis no such paradox arises. The following derivations show 
the interaction of schwa syncope and nasalization in forms that proved 
problematic for Narang and Becker: 


a:gsno: 'courtyard' 


da:nwo: 'demon' 

K I K 


syllabification (16) 
and foot const. (18) 

schwa desyll. (20) : 

core syllab. (16) : 
schwa reinst. (27) : 
core syllab. (16) : 





I i I 



V \ 







da nawon 

^a A6w6n 


I 1/ I |l ; 

da nSwoii 


da nawon 

The schwa facts have now been accounted for. The nasalization facts can 
be accounted for as follows: 


nasal spread (48) 

X X X X X X X 
I J I I I I I 

/K-. .'K 

k n g ^ n 

d a 

[ ] [ ] 

[ ] [ ] 
+N I 

n 3 w o n 





core syllab. (16) 
vowel spread (49) 

X X X X x' X X 

xxxxxxx X 

III:;: 1 

a n g a n o n 

da n a w o n 

k> \/^ 


[ ][ ] [ ][ ] 

+N ' +N 1 

[ ][ ] 

+N \ 





core syllable (16): 

PR a:gano: da:nwo: 

No matter liow the rules are ordered, the correct results are obtained 
because the coda position is filled in underlying structure and so 
prevents core syllabification from incorporating any other element into 
the coda. This analysis crucially depends on the distinction between 
solid lines and broken lines which allows the differentiation of elements 
incorporated into the syllable by rules and those present in underlying 
representation. Without this distinction it would not be possible to 
differentiate between nasals which are a part of the syllable in UR and 
affected by nasal spread and those incorporated into the syllable by rule 
and immune to nasal spread. 

The rule ordering necessary to account for the e/0 alternation in 
Dialects A and B makes certain predictions about a/0 alternation after 
nasal vowels, namely that in Dialect A schwa should not delete when 
preceded by a nasal vowel while in Dialect B schwa should delete in this 
environment. In fact these predictions are borne out and in Dialect B 
one finds alternations like the following: 

(52.) a:gano: ■-~'a:gno: a:calo:'-' a:clo: 

while in Dialect A *a:gno:, *a:clo:. We have already seen a sample 
derivation from Dialect A; given below is the derivation of the word 
a : gno : in Dialect B. 

(53.) after schwa XXXXXp 

desyllab. (20): 

core syllab. (16) : 
extended syllab. (30): 

schwa reinst. (27) : 

^ \ 






1 1 ' ' ' ' I 

(Nasal and vowel spread are not affected by any of these processes 
because they take place on a different plane.) 

Thus an additional benefit accruing from the treatment of nasal 
vowels as a VC sequence is that it makes possible a single unified 
account of both extended applications of schwa syncope. Narang and 
Becker, however, were unable to reap this benefit in their linear 
analysis . 

4.0 Conclusion. 

What I have sought to demonstrate is that a syllable based 
approach to phonological processes, schwa syncope and vowel nasali- 
zation in this instance, provides a relatively simple, motivated account 


of the phenomena. It accounts for all the Hindi-Urdu data with a few rules 
most of which are independently needed in the language. I have also 
shown that the nasal vowels in Hindi-Urdu can be treated as a sequence of 
vowel and consonant thus achieving a considerable economy in the phoneme 
inventory of the language. Vowel length is easily explained, and I have shown 
that syllable structure accounts for facts that seem arbitrary stipulations 
in a linear analysis and eliminates the need for global rules. The latter is 
an important consequence of non- linear phonology and makes a case for the 
need for several planes of representation so that information crucial to a 
derivation is available on some plane at every point in the derivation. 
Linear phonology can only get this information by adding power to the 
grammar whereas non- linear phonology gets it as a natural consequence of 
the theory. The implications of this are worth investigating. 

Of theoretical interest is the fact that this segment of Hindi- 
Urdu phonology can be dealt with by means of an X skeleton and a fairly 
straightforward structural distinction between +N and -N elements. No 
need was found for CV labels (Clements and Keyser 1983) or for a 
hierarchical structure as proposed by Levin (1984). Of course this 
paper looks at but a small part of the phonology of Hindi-Urdu, however, 
the fact that this complex little area does not require CV labels or an 
elaboration of structure is of some significance. 

This paper has benefited greatly from discussions with Diana 
Archangeli. Thanks are also due to Hyang Sook Sohn for her valuable 
comments, and to Yamuna Kachru and Rajeshwari Pandharipande for help 
with the data. Any faults are of course mine. 

The language described here is Hindi-Urdu or Standard Hindi. 

Hindi-Urdu is the language used by educated native speakers in every- 
day conversation. 

denotes length. 

cription; I have standardized all forms to provide uniformity. 

In H-U the long vowels are [+tense] the short vowels are 
[-tense] . 

Clements and Keyser formulate the principle as follows: 
"Syllable-initial consonants are maximized to the extent consistent 
with the syllable structure conditions of the language in question." 

According to Archangeli (1984) core syllabification rules apply 

automatically after the application of any rule. The consonant stranded 

by schwa deletion is thus automatically taken into an adjoining 

syllable, in this case the preceding one since it has no coda and the 

following syllable already has an onset. 


I am grateful to Diana Archangel! for bringing this to my 

Assuming that all long nasal vowels are underlyingly short 

enables one to account for the fact that on the surface one finds very 

few short nasal vowels but a plethora of long ones. The few forms with 

short nasal vowels can be treated as lexically marked exceptions. 

This is not an unmotivated move for as Ohala (1972) observes, 

"Interestingly since the short vowel forms go against the general 

tendency of the language, many times the nasalization is lost, i.e. nati 

speakers seem to treat these as exceptions." 

It is here that the advantage of an X skeleton becomes obvious. 
If a CV skeleton had been used one would have to link a vowel to 
a C slot, not a very desirable move. 

The nasalization facts can also be dealt with if we assume a 
biplanar representation of features e.g. 

[+nas] [+nas] 

X X Rules 48 and 49 apply 




I J 

L J 





[ ] 



[ ] [ ] 


a uniplanar representation however would give the wrong results, 
unless one added a stipulation to the effect that all features but 
nasal delete e.g. 



ARCHANGELI, D. 1984. Underspecif ication in Yawelmani phonology and 

morphology. Doctoral Dissertation. MIT. 
BHATIA, T. and M. Kenstowicz. 1972. Nasalization in Hindi: a 

reconsideration. Papers in Linguistics 5:2. 
CLEMENTS, G. and S. J. Keyser. 1983. CV phonology: a generative theory 

of the syllable. Linguistic Inquiry Monograph 9. MIT. 
D'SOUZA, J. 1983. Syllable structure in Hindi. Unpublished ms. 

University of Illinois. 
GOLDSMITH, J. 1976. Autosegmental phonology. Doctoral Dissertation. 

MIT. Dist. lULC. 
GROVER, H. n.d. Arabic root and pattern morphology without tiers. 

Unpublished ms. Michigan State University. 
HAMMOND, M. 1984. Constraining metrical theory: a modular theory of 

rhythm and destressing. Doctoral Dissertation. UCLA. Dist. lULC. 
. 1985. Main stress and parallel metrical planes. To appear 

in the proceedings of BLS II. 


KENSTOWICZ, M. and C. Kisseberth. 1979. Generative phonology. New 

York: Academic Press. 
KIPARSKY, P. 1983. Abstractness , opacity and global rules. In O. 

Fujimura, ed. : Three dimensions of linguistic theory. Tokyo: TEC. 
LEVIN, J. 1984. Conditions on syllable structure and categories in 

Klamath phonology. To appear in the proceedings of the West 

coast conference on formal linguistics. 
MCCARTHY, J. 1981. A prosodic theory of nonconcatenative morphology. 

Linguistic Inquiry 12. 
NARANG, G. C. and D. A. Becker. 1971. Aspiration and nasalization in the 

generative phonology of Hindi-Urdu. Language 47:3. 
OHALA, M. 1972. Topics in Hindi-Urdu phonology. Doctoral Dissertation. 

PRAY, B. 1970. Topics in Hindi-Urdu grammar. Research Monograph Series 

No. 1. Center for South and Southeast Asia Studies. University of 

California, Berkeley. 
PULLEYBLANK, D. 1983. Tone in lexical phonology. Doctoral Dissertation. 

RAPPAPORT, M. 1984. Stress and ultra-short vowels in Tiberian Hebrew. 

To appear in the proceedings of the West coast conference on 

formal linguistics. 
STERIADE, D. 1982. Greek prosodies and the nature of syllabification. 

Doctoral Dissertation. MIT. 
WITHGOTT, M. 1982. Segmental evidence for phonological constitutents . 

Doctoral Dissertation. University of Texas, Austin. 

studies in the LirnQjistic Sciences 
Volume 15, Number 1, Spring 1985 


Andrea S. Dunn 

This paper describes the structure and activities of the National 
Swahili Council (BAKITA), the official governmental agency in 
charge of the coordination and implementation of Tanzania's 
Swahili lanugage policy. One of BAKITA's activities, the 
preparation of new vocabulary for subjects where Swahili hsis not 
previously been used, has been criticized in preliminary 
evaluations which show that the new vocabulary is not readily 
accepted. It is suggested here that problems in the creation, 
dissemination, and acceptance of new vocabulary might be avoided 
through the adoption of a model of the vocabulary expansion 
process such as that proposed by Marshad (1984). 

Following Kloss' (1969) definition, language planning may be divided 
into status planning and corpus planning. The former set of activities 
includes the allocation of languages to various functions within the 
community (often nation), and is usually the result of policy making. 
Corpus planning includes the standardization, elaboration and codification 
activities necessary to enable the languages to fulfill their designated 
functions. While these two sets of language planning activities are often 
carried out by different organizational structures, the basic claim that 
this paper demonstrates is that a model of corpus planning cemnot be 
constructed without a clear, well supported language policy. 

Few studies have appeared on the organizational structure and actual 
operation of specific language (=corpus) plemning agencies (Fellman and 
Fishman, 1977). Fellman and Fishman (1977), for example, give a summary of 
the activities and organization of two committees of the Hebrew Language 
Academy. Their goal is to examine the actual functioning of the language 
planning process in order to eventually include the workings of the language 
planning bodies (their meiii>ership , structure, work habits, etc.) as 
variables for study and evaluation in the language (=corpus) planning 
process. In the case of Israel, however, the modernization of Hebrew 
received full governmental and institutional backing. Furthermore, the high 
level of education and literacy of Israeli citizens provided a foundation 
for the development of modern Hebrew. 

The purpose of this paper is to examine the workings of another 
language planning agency, The National Swahili Council (BAKITA), which 
operates in a different sociolinguistic environment them the Hebrew Language 
Academy. Although in many ways the processes and problems facing vocabulary 
planners are the same everywhere, the study of BAKITA shows that the overall 
success of such organizations depends on their having a model of the 
vocabulary expansion process that provides for a realistic view of the 
sociolinguistic situation (users/audience etc). Such a model cannot be 
formed without a clear language policy. This was lacking in Tanzania, 


especially in the educational sector. Now that this has been clarified, 
however, language planners in Tanzania are better equipped to construct a 
model of the vocabulary expansion process. 

In this paper I will describe the structure and operation of BAKITA, 
focusing on problems currently under treatment. First I will present a 
brief summary of contemporary Tanzanian language policy and the history of 
Standard Swahili in order to put the work of BAKITA into perspective. The 
main part of the paper is divided into two sections. The first describes 
the organizational structure, membership and duties of BAKITA. The second 
provides examples of problems currently being dealt with at BAKITA. The 
paper concludes with observations on the organization's success in achieving 
its goals. Recommendations are made, in light of recent modifications in 
educational language policy, for improving Bakita's effectiveness in 
implementing the policy through the adoption of a model of vocabulary 


The United Republic of Tanzania, located on the East African Indian 
Ocean coast between Kenya to the north and Mozambique to the south, has a 
population of approximately 19 million people who speak over 120 languages. 
Languages belonging to all the major African language families are spoken in 
Tanzania (Nilotic, Khoisan, Afro-Asiatic and Niger-Congo). However, the 
majority of Tanzanians are speakers of Bantu languages, one of the branches 
of the Niger-Congo family (Polome', 1980). Two languages, English and 
Swahili, serve as lingua francas, though their functional allocation is 
quite different. English, taught as a school subject and used as the medium 
of instruction in post-primary education, is known by a very small 
proportion of the population. Swahili, on the other hand, as the national 
amd official language, serves as the medium of primary education, and is the 
language of wider communication. Swahili is known by well over 90% of the 
population. 2 

Swahili spread as a lingua franca throughout East Africa, but 
especially in Tanganyika, shortly before and during the colonial era.^ Its 
spread was accelerated by the necessities of colonization, world wars, and 
the independence movement, during which it became the medium of national 
unity (Abdulaziz, 1972a). Shortly after Independence in 1960, Swahili was 
declared the national and official language. In 1967, it was declared the 
medium of instruction in all primary schools as well as the required medium 
for all official government communication. Also in 1967, the Baraza la 
Kiswahili la Taifa - BAKITA - (the National Swahili Council) was formed to 
implement Tanzania's official Swahili language policy." 

Language policy in Tanzania refers overtly to the use of Swahili and 
English in two sectors of national life, government and education (Dunn, 
1984). The policy has evolved since Independence through the statements of 
individuals in policy-making positions. According to the policy, Swahili is 
to be used as the medium of participation in national affairs at all levels 
and in all branches of the government, though in practice, in a few areas 
such as at the higher levels of the legal system, English is used. Swahili 
is also the medium of primary education and primary teacher training. 
Policy formulated in the late 1960s and early 1970s called for the switch 


from English to Swahili as the medium of instruction at the secondary, and 
eventually at the tertiary, levels. However, following a long period of 
uncertainty, this policy was officially reversed in 1983 by the Minister of 
Education and in 1984 by the President. Speaking at the opening of a 
seminar of UKUTA (Usanifu wa Kiswahili na Ushairi Tanzania - Tanzanigin 
Society for Standardization of Swahili and Poetry) President Nyerere 
confirmed that the policy had been changed. 

'English is the Swahili of the world and for this reason must be 
taught in [Tanzania] ...'... The President added that English will 
indeed be the medium of instruction in secondary schools and 
institutions of higher education. ( Mzalendo , October 28, 
1984 ).s 

In the governmental sector, where the Swahili language policy has been 
clear and well articulated, implementation proceeded well. In the 
educational sector, on the other hand, the vacillation among policy makers 
as to whether the go-ahead would be given for the use of Swahili as the 
medium of secondary education caused a certain amount of uncertainty among 
language planners. While vocabulary lists and a number of Swahili textbooks 
were prepared in anticipation of the switch-over to Swahili medium, without 
a clear sense of whether Swahili would actually be allocated for use in this 
function, work did not proceed at a rapid pace. Since 1983, however, the 
status of Swahili in the educational sector seems clear, and the task of 
corpus planners, therefore, better defined than in the past. Since the 
policy seems to limit the role of Swadiili in the educational sector to 
primary and adult education, BAKITA can coordinate the task of vocabulary 
expansion with these audiences in mind, producing vocabulary which will be 
acceptable to them. 

Kiswahili Sanifu (Standard Swahili) is the variety of Swahili which is 
referred to in Tanzania's language policy and whose correct use and 
elaboration is of concern to BAKITA. Standard Swahili formally came into 
being in the 1930s when the members of the Inter-territorial (SwaJiili) 
Language Committee selected the dialect of Swahili spoken in Unguja, the 
main town of the island of Unguja (Zanzibar), as the basis for the standard 
language (Whiteley, 1957, 1969). This committee, composed of European 
representatives from each of the territories of East Africa, sought to 
normalize the use of Swahili, particularly in education, in order to make 
possible the use of standard textbooks and examinations throughout the 
territory. For a short time, the committee held an imprimatur on the 
publication of all Swahili books, but this control was lost with the 
expansion of Swahili publishing after World War II. The committee's early 
work also included the revision of the Swahili/English, and the production 
of the English/Swahili dictionaries (Interterritorial Committee, 1939) which 
served virtually as the only dictionaries of the language until the 
publication in 1980 of the Kamusi ya Kiswahili Semifu (Standard Swahili 
Dictionary) (Institute for Kiswahili Research). 

After World War II, the committee, now comprised of European and 
African members, retained its authority in matters of Swahili usage, 
although exercise of this authority was limited due to the static budget of 
the committee and the creation of the East African Literature Bureau which 
oversaw Swahili publications and was not obliged to seek the imprimatur from 


the committee (Whiteley, 1969). Most of the connnittee's attention during 
the last decade before Independence was devoted to the collection of 
traditional Swahili manuscripts and to the production of a series of 
monographs on the Swahili dialects. Throughout its life the committee 
published its Bulletin , later called Journal and later Swahili (now 
Kiswahili ) , which forms a valuable repository of studies on Swahili language 
and literature. 

In the decade after Independence, the duties of the committee were 
distributed among the two language treatment agencies currently operating 
with government funds in Tanzania. The Taasisi ya Uchunguzi wa Kiswahili 
(the Institute for Kiswahili Research - TUKI) formed in 1964 and located at 
the University of Dar es Salaam, is the direct descendant of the committee. 
With its staff of linguists euid language scholars, TUKI is primarily 
responsible for Swahili research and lexicography, that is the academic 
aspects of the implementation of Tanzania's Swahili language policy. 
BAKITA, created in 1967, inherited the authority that the committee 
exercised to establish standard Swahili usage, including supervising the 
expansion of vocabulary in technical fields. This two-pronged approach to 
the development of Swahili creates a situation of conflict of authority 
which has undoubtedly had an effect on the implementation of the nation's 
Swahili language policy. 

Structure of BAKITA 

BAKITA is the official body charged with coordinating the 
implementation of Swahili language policy in Tanzania. Formed by an act of 
Parliament in 1967, the duties of BAKITA are similar to those of the former 
Inter-territorial (Swahili) Language Committee (Whiteley, 1957, 1969). 
BAKITA seeks to further the development and use of Swahili in Tanzania; to 
motivate the use of Swahili in all spheres of public life; to encourage 
cooperation with other groups concerned with the development of Swahili; to 
prescribe correct use of Swahili and to prevent incorrect use; to oversee 
the translation of technical terms, to produce journals and pamphlets 
concerning the correct use of Swahili; to serve the government and public 
corporations as well as private individuals in solving problems which arise 
from the use of Swahili in their activities (Irira, 1983). As a result of a 
1983 amendment to the law forming BAKITA, which enables the organization to 
establish international contacts, the organization's goals have been 
expanded to include cooperation with neighboring East African countries in 
the promotion and development of Swahili, as well as with those involved in 
the teaching of Swahili as a foreign language in Tanzania and abroad. 

BAKITA, eis a parastatal organization, is a semi-autonomous governmental 
body under the policy direction of, and financially dependent upon, the 
ministry to which it is attached. In the past BAKITA has been attached to 
the ministries of Regional Administration and Rural Development (1968), 
National Education (1969-1973) and Youth and National Culture (1973-1979). 
BAKITA has been part of the Ministry of Information and Culture until its 
reorganization as part of the Prime Minister's office in 1984. 

Internally, BAKITA consists of two main divisions, the Baraza (council) 
and the secretariat. The Baraza, whose chairman is appointed by the 
President, is the authoritative body of BAKITA. The full Baraza meets two 


times per year as do each of its committees: standardization, translation, 
editing and publication, and grammar and imprimatur. Its 53-person 
membership is selected from every region and language-related institution in 
the country, and has final authority on decisions affecting Swahili. 
Delegates to the Baraza are chosen on the basis of their interest and 
experience, professional or non-professional, in the Swahili language, not 
necessarily for their expertise in the formal study of language. The 
delegates to the Baraza are appointed by the minister to whose ministry 
BAKITA is attached based upon the recommendations of the executive committee 
of BAKITA. All delegates to the Baraza serve a concurrent three year term 
after which the Baraza is dissolved and reconstituted. 

The Secretariat of BAKITA, which consists of 24 full-time employees 
responsible for the day-to-day activities necessary to prepare work for the 
Baraza' s approval, is divided into departments which correspond to the 
committees of the Baraza: standardization, translation, editing and 
publication, and grammar and imprimatur. Eleven of the 24 employees are 
academic staff with training to the Bachelor's degree level in a 
language-related field, though not in terminology, lexicology or language 
planning. In addition, the secretariat also has the department of finance 
and management, and the office of the executive secretary which oversees the 
functioning of the organization. 

The academic employees of BAKITA are those who do the research amd 
coordination of efforts involved in developing and standardizing Swahili. 
They are present but do not vote at the meetings of the Baraza and its 
committees, where one member from each department serves as secretary of the 
committee to which his/her department is responsible. The academic 
employees also act as representatives of the Baraza in various situations 
where issues of language use arise. The disbursement of funds is 
controlled, and the calendar for the Baraza is planned, by the executive 
committee which consists of the executive secretary and several members of 
the Baraza. The chairman of the Baraza is also chairman of the executive 

The work of BAKITA' s committees is summarized below; issues arising 
from the work are discussed in the following section of the paper. 

The committee/department on language standardization is primarily 
responsible for setting guidelines for and for overseeing lexical 
elaboration. The department prepares and disseminates lists of terms 
prepared by their members of staff in cooperation with other agencies. 
Inquiries from writers concerning vocabulary are directed to this 
department. This department also cooperates with the department/committee 
on grammar and imprimatur in resolving problems of usage which arise in the 
vocabulary expansion process. The results of the department's efforts are 
approved by the committee on standardization of the Baraza, as well as the 
full Baraza, after which they may appear in Tafsiri Sanifu (Standard 
Translations), the list of official translations. The department of 
language standardization also prepares monolingual classified vocabularies; 
however, none of these have been published due to budgetary restrictions. 

The grammar and imprimatur committee/department is concerned with 
problems of correct usage. This department conducts research into 


grammatical problems arising from the vocabulary expansion process as well 
as into trends in usage. The department also attempts to control usage in 
the media by occasional monitoring of the radio and newspaper. Manuscripts 
for school texts and other books are brought to this department for approval 
of the language before being sent for publication. The decisions on usage 
reached by this department/committee are disseminated in the pamphlet series 
Jifunze Kiswahili Uwafunze Wengine (Learn Swahili to Teach Others). 

The department/committee on translation is divided into two sections, 
the first deals with the translation from European languages into Swahili, 
and the second with translation into Swahili from the ethnic group languages 
of Tanzania. The department was created in 1977 at the request of the Ghana 
cha Mapinduzi (the sole political party) in order to fill the need for books 
in Swahili (Sambazi, 1984). At that time, a list of 25 books on politics, 
economics and development was suggested for translation. A number of these 
have been or are in the process of being translated. In addition, the 
department also translates school texts and reference books as well as forms 
and pamphlets from government departments, public corporations, government 
ministries and private individuals. The translation department also 
translates written and oral literature from other Tanzanian languages into 
Swahili as a way both to preserve traditions and to enrich the vocabulary of 
Swahili by borrowings from ethnic group languages. 

The editing and publications department/committee is concerned with the 
dissemination of the views and decisions of BAKITA via publications and the 
mass media. The weekly radio program Lugha ya Taifa (National Language) 
deals with efforts by people around the country to promote the use of 
Swahili as well as with the work of BAKITA. The journal Lugha Yetu (Our 
Language) which appears two times per year carries articles written in 
Swahili on introductory topics in linguistics, correct usage, poetry and 
other forms of Swahili literature, as well as lists of standard 
translations. Tafsiri Sanifu (Standard Translations) provides bilingual 
English- Swahili lists of equivalents which have been approved by the Baraza. 
The pamphlet Jifiinze Kiswahili Uwafunze Wengine (Learn Swahili To Teach 
Others), prepared by the department of grammar and imprimatur, gives points 
on correct Swahili usage such as distinguishing often confused pairs of 
words. This department also cooperates in the annual book display, staged 
at the national trade fair. 

Activities of BAKITA 

Since its inception in 1967, BAKITA has grown from a one-member staff, 
forced to share office space with the National Youth Organization, to a 
staff of over 20 members, occupying their own office space in the center of 
Dar es Salaam. The organization's early work concentrated on implementing 
the use of Swahili in the governmental sector through activities designed to 
raise the status of Swahili and persuade all members of the public sector to 
use Swahili in their offical communications (BAKITA, 1969-1981). In order 
to assist in this endeavor, BAKITA coordinated the preparation of lists of 
Swahili words for office and factory names and titles, as well as the 
translation of forms, oaths, receipts and other paraphernalia of 
bureaucracy. Once Swahili was firmly established in the governmental arena, 
BAKITA turned its attention to the preparation of vocabulary to make Swahili 
an adequate instructional medium for all school subjects. With an expanded 


staff, BAKITA has focused in recent years on lexical elaboration and the 
problems arising from this task, while continuing to prescribe Standard 
Swahili usage. 

Prescribing Usage 

The definition of correct usage is an on-going process, particularly in 
the sociol inguist ic situation in which Swahili is found. Despite the fact 
that nearly all Tanzanians speak Swahili, many speak it as a second language 
and their competence in it varies depending on such factors as urban or 
rural living environment, degree of proximity to the coast, degree of 
monolingualism among speakers of the ethnic group (first) language, extent 
of opportunities for travel, and degree of formal education. Knowledge of 
Standard Swahili is favored by opportunities for schooling beyond the third 
grade level, for travel and work outside the ethnic group area, and by urban 
living. In such a setting, variation in Swahili is to be expected. 
BAKITA's staff proposes standard usage to normalize this variation. Recent 
standardization activities have included development of a standard style 
sheet for use by writers and editors, a call for the elimination of the 
influence of English on Swahili, regularization of the orthography, and 
clarification of isolated cases of poor usage (e.g., confusion of near 
synonyms, incorrect plural izat ion, or incorrect agreement). 

The pamphlet Jifunze Kiswahili Uwafunze Wengine (Learn Swahili to Teach 
Others, BAKITA, 1983), a guide to correct Swahili usage, warns against the 
influence of English. This influence occurs in several forms from the overt 
mixing of English and Swahili (Waomba.)i wawe na license class C 
'Applicants should have a class C license ' instead of Waomba.ji wawe na 
leseni dara.ja la Ch) to the translation of English idioms into Swahili 
(kukamata bas i 'to catch a bus' instead of k upanda basi 'to board a bus'). 
The influence of English is also seen in the abbreviations of the names of 
public corporations. While the name in its full form is given in Swahili, 
the commonly used abbreviation, even when written in Swahili, is based on 
the former English name, a practice which BAKITA is attempting to change. 
Thus, for example, the commonly used abbreviation of Benki ya Biashara ya 
Taifa is NBC , from the English name National Bank of Commerce . Similarly, 
University of Par es Salaam is abbreviated UDSM, rather than CKDSM, the 
abbreviation of the Swahili translation Chuo Kikuu cha Par es Salaam . 

Currently, BAKITA is also suggesting orthographic changes which would 
effect a very small number of words in the lexicon. While in most cases 
Swahili orthography is extremely phonemic, in a few cases the syllabicity of 
an initial nasal, which is not indicated by the current orthography, serves 
to distinguish two words. For example /mbu-ni/ (two syllables) 'ostrich' 
and /m-bu-ni/ (three syllables) 'coffee plant' are both spelled mbuni . 
Similarly, /mba-ya/ (two syllables) 'bad, modifying class 9/10' and 
/m-bay-a/ (three syllables) "bad, modifying class 1 or 3' are both spelled 
mbaya . In other cases, although no homographs exist, the spelling does not 
indicate the status of the nasal. Thus words such as nje "outside', nge 
"scorpion', or mbwa "dog', give no clue to the fact that their nasals are 
syllabic. The failure to indicate the status of the nasal is felt by BAKITA 
and some Ismguage teachers to pose problems for learners of Swahili. 

In order to rectify this perceived deficiency, BAKITA has proposed to 


indicate a syllabic nasal by either doubling the nasal or by separating the 
nasal from the following consonant with an apostrophe. According to the 
first proposal, words with syllabic nasals would be written with a double 
consonant as in mmbuni 'coffee plant*, or 'outside'. Words with 
non-syllabic initial nasals would be written following the former spelling; 
mbuni 'ostrich' and mbaya 'bad, modifying class 9/10'. The second proposal 
involves the use of an apostrophe to indicate syllabicity. Thus 'coffee 
plant' and 'outside' would be spelled m'buni and n' je respectively. 
Non-syllabic nasals would retain the original spelling. At present BAKITA 
has called for more research into the need for such orthographic changes. 
Before any changes are decided upon by the Baraza, the organization will 
seek the full cooperation of Swahili experts as well as input from the 

Lexical Elaboration 

By virtue of its very nature as a body dedicated to the development and 
promotion of Standard Swahili, BAKITA's activities are fundamentally 
prescriptive. However, BAKITA also assumes an elaborative role in order to 
extend the use of Swahili to areas where it had not previously been used, 
for instance in publications for the general public on technical topics, 
such as photography, knitting, nutrition, shorthand, carpentry, etc, or in 
the preparation of materials for the upper primary school subjects. In this 
task BAKITA works with other organizations and individuals to expand the 
vocabulary of Swahili in order to fill the lexical gaps felt primarily by 
writers as they write on subjects in Swahili for the first time. In 
concerning itself with lexical elaboration BAKITA is fulfilling the purpose 
for which it was created, the development of Swahili. At the same time, 
BAKITA is attempting to mediate the influx of new words into the language by 
establishing guidelines for their acceptance. 

Guidelines for Vocabulary Expansion 

In their discussions, delegates to the standardization committee 
meetings attempt to follow criteria for the source and structure of the 
words to be standardized (BAKITA, n.d.; Irira, 1982). Criteria concerning 
the source for borrowing specify that vocabulary items are to be taken from 
the following languages or groups of languages in order of preference: 
Swahili; other Tanzanian Bantu languages; other African languages (excluding 
Arabic); foreign languages. Words taken from Swahili may be taken from any 
of its dialects ( kisabeho 'breakfast' < Kipemba), or they may be 
morphological derivations on existing roots ( kitenguo 'derivative* 

< kutengua 'to separate*). Existing Swahili words are also used to form new 
words by compounding ( pembenane 'octagon' < pembe 'corner' and nane 
'eight*) or by the shortening of nominal phrases ( kiuavi jasumu 'antibiotic' 

< kitu cha kuua vijidudu vya sumu 'something to kill very tiny poisonous 
insects'). The semantically opaque words obtained by this method, however, 
are often not acceptable to members of the standardization committee. 

While other Bantu languages may serve as sources for vocabulary 
expansion in Swahili, in practice, there is little systematic surveying of 
these languages for words that would fill the gaps that have been 
identified, partly due to a lack of reference materials for these languages, 
and partly due to the fact that these languages are not used in the subjects 


for which words are sought (as is English, for example). Words from ethnic 
group languages most often find their way into standard Swahili through the 
presence, at some stage of the vocabulary translation and staindardization 
process, of a speaker of that language who suggests that his or her language 
has a suitable word for the concept under discussion. After the meaning and 
context of use of the word is explained to other members, and the committee 
is satisfied, the word is adopted (e.g., fuwele 'crystal' < Kizigua, 'a 
soft, breakable stone', or ikulu 'statehouse' < Kinyamwezi/Kisukuraa) . Much 
the same procedure applies to the use of non-Tanzeinian African languages as 
a source for vocabulary expansion. In this case, as there are few 
non-Tanzanian Africans present during the vocabulary expansion process, few 
words from other African languages are adopted.^ 

English is clearly the first non-African foreign language source from 
which vocabulary is borrowed. Arabic, an African language, and the main 
source for borrowing in the past, is now stated to be in the least preferred 
position; however, in practice which language acts as a source depends on 
the subject for which vocabulary is sought. The strength of English as a 
source, for example, is felt most in technical fields, while Arabic words 
are more often used in the humanities. Thus, of the 121 borrowed words in 
the list of 349 biology terms (BAKITA, 1980), all are from English. In 
contrast, in a list of 243 literature terms (BAKITA, forthcoming) only five 
are borrowed from English while perahps half are words of Arabic origin, 
most of which are semantic and/or morphological extensions of nativized 
words. This difference, of course, reflects the dynamics of the contact 
situation in which Swahili finds itself: English, as the medium of higher 
education, is the language of the technical registers. On the other hand, 
since Arabic has contributed for a long time to the development of Swahili 
culture, it is not surprising that words needed for a register of literary 
criticism should come from the Arabic part of the Swahili lexicon or from 
Arabic itself. 

In addition to specifying the source from which vocabulary expansion 
may draw, BAKITA also attempts to control the structure of the words and 
their integration into the grammatical system of the language. Guidance in 
the first task is achieved by prescribing rules for the nativization of 
borrowed words. The department of standardization has a policy of 
nativization of borrowed words which states that all words, except names of 
individuals, will be modified phonologically to conform to the sound system 
of Swahili by the addition of a final vowel for words which are consonamt 
final in the source language, and by the insertion of vowels to breeJt up 
impermissible consonant clusters (Irira, 1983). Overly long words, created 
by compounding, are also discouraged. 

As a result of these guidelines, words ending in m have a u added: 
sodium- sodiamu , ileum- ilamu , calcium- kalisiamu . Words ending in other 
consonants have an i added: protein- protini , auricle- orikali , 
cellulose- sellulosi . Vowels are introduced word-internally in a number of 
cases to break up consonant clusters which are not permitted in Swahili. So 
calcium is rendered as kalisiamu and cytoplasm as sitoplazimu . In other 
cases, consonant clusters are simplified by the deletion of one or more of 
the members: organ- ogani . artery- ateri , thermometer- them ometa , 
metabolism- metaboli . As these examples show, vowel phonemes may also be 
modified in the process of nativization in order to make them conform more 


closely to the Swahili sound system although no guidelines for these 
modifications are provided by BAKITA. 

Unfortunately it is not always possible in the standardization process 
to apply these criteria regularly and rigidly, because members of the 
standardization committee and the Baraza are also guided by general, 
abstract principles about the ideal nature of the vocabulary which they are 
creating. These, sometimes contradictory, principles include consistency, 
transparency, precision, cultural synchronicity and intertranslatability . 
Consistency is an oft-stated goal as committee members call for terms to be 
created following particular morphological patterns. Thus committee members 
objected to the contraction of phrases into single words as a method for 
vocabulary expansion (e.g., kiuavi.jasumu ) because no regular pattern, such 
as retaining only the first syllable in the word, would produce acceptable 
terms. In amother statement of the need for consistency, MacWilliam, a 
researcher at the Institute of Kiswahili Research, suggests that a single 
Swahili morpheme be used for all occurrences of a particular Latin or Greek 
affix in the preparation of technical terminology (MacWilliam, 1983). 
MacWilliam notes that the affix iso- , for example, is rendered 
inconsistently in Swahili as iso , also , or by the Swahili word sawa 'equal'. 
She proposes that one of these alternatives be approved by BAKITA and used 
whenever translations are sought for terms containing the affix iso- . Mdee 
(1983) also calls for consistency in BAKITA' s adaptation of loan words to 
the phonological system of Swahili. 

Committee members also express the belief that words and terms should 
be transparent, that is, that the meaning of the whole should be readily 
deducible from the parts. This principle implies that, for most Tanzanian 
audiences, words and terms should be created by morphological derivation or 
compounding on existing Swahili roots. This principle, therefore, militates 
against the contraction of Swahili phrases into opaque words, or the 
wholesale borrowing from other languages. Vocabulary expansion by semantic 
extension is also not favored as it is thought that users will not be able 
to distinguish the "new" or technical use of a word from its common use and 
will, as a result, not realize when a word is used in its technical sense. 

The principle of transparency, however, often contradicts another 
principle, precision. Members of the standardization committee and the 
Baraza feel that coined terms must accurately reflect the concept which they 
represent and be satisfactory for specialists in the field. This principle, 
unlike that of transparency, favors borrowing of terms, particularly from 
the languages in which the concepts were first learned by current 
practitioners (usually English). In a move towards more precise 
terminology, several terms based on Swahili roots which had been approved 
earlier and were in common use were replaced by nativized English loans. 
Thus, wanga 'starch' and , formed from a contraction of hamira 
'yeast' and 'starchy gravy', both of which were used as equivalents for 
'carbohydrate' , were replaced by the nativized English loan, kabohaidreti . 

In preparation of terminology for technical subjects, the goal of 
precision often implies the nativization of loan words, a trend which runs 
counter to the principle of cultural synchronicity. Tanzanians are well 
aware of the unifying force and symbolic significance of their national 
language. Elaboration of vocabulary in Swahili is seen as a patriotic 


contribution to the development of the nation. In this context, therefore, 
it is not surprising that a number of members of the standardization 
committee and the Baraza as a whole stress the necessity of using' Swahili 
roots as the basis for vocabulary expansion in order to preserve the 
national character of the language. At the same time, however, other 
Tanzanians who feel strongly the need to preserve intertranslatability in 
the language of science and technology, advocate the use of English or 
internationally recognized standard roots as a basis for vocabulary 
expansion, particularly in the technical registers. 

The contradiction between these principles are part of the dynamics of 
the vocabulary expansion process and, in part, explain why rigid guidelines 
for the preparation of vocabulary are difficult, to formulate and impossible 
to follow. However, a number of the contradictions could be resolved if 
vocabulary were prepared with specific audiences of potential users in mind. 
That is, while the principles of cultural synchronicity and transparency are 
important when formulating guidelines for the preparation of vocabulary for 
use in lower primary school and in adult literacy and literacy maintenance, 
the principles of precision and intertranslatability are more important for 
upper-primary school and primary teacher training. Awareness of the 
different needs of different audiences, and hence the use of different 
guidelines for preparation of vocabulary appropriate to each, would enable 
BAKITA to produce a more consistent, acceptable product. 

Procedures for Vocabulary Expansion 

While some translations of vocabulary are prepared by BAKITA employees, 
most are the result of a joint effort by subject area specialists, who are 
the users of the vocabulary, and BAKITA, via the language standardization 
department/committee. Subject specialists such as curriculum developers, 
textbook authors and teachers translate the words which they need. These 
words are then brought to BAKITA where they are scrutinized according to the 
guidelines prepared by the language standardization department. After this, 
the words are discussed by the language standardization committee in one of 
its semi-annual meetings. These meetings, which last for ten days to two 
weeks, are held in various regional towns. Participsmts include the head of 
the language standardization department in the role of secretary, delegates 
of the Baraza, as well as members of the public such as school teachers from 
the area in which the meeting is held. The meeting is conducted by one of 
the delegates who is selected by the voting members present. Although such 
a large and varied panel may be unwieldy, it provides public input into the 
vocabulary expansion process which helps standard Swahili achieve a national 
character, and leads to its acceptance.'' 

The process by which the words are accepted as standard Swahili is 
often time-consuming. Each word is discussed separately, and the concept on 
which it is based is elucidated. Competing words and their reasons for 
rejection are mentioned as are related words which have already been 
standardized. If specialists in the subject are present, who may be the 
authors of the words, they are asked to explain the concept for which the 
word is sought and the rationale for the proposed choice. All participants 
in the meeting are free to voice their objections and propose alternatives. 
Decisions on whether to adopt words are not usually postponed and discussion 
continues until a consensus is reached. 


The discussion preceding the adoption of the word ngaha 'accessories' 
is illustrative of the procedure followed by the committee on 
standardization. The original word proposed by the literature department of 
the Institute of Kiswahili Research, who prepared the list of terms for the 
field of literary criticism, was visovasi , a contraction of the phrase vitu 
ambavyo si vazi 'things which are not clothes', with reference to theatrical 
properties used by actors to enhance their costumes but which are not the 
actual costume/clothes. Delegates objected to visovasi on the grounds that 
it implies 'things which are not worn', since the verb from which the noun 
vazi "clothes' is derived, kuyaa, 'to wear, put on' is also the root for the 
last part of the phrase vasi. Delegates also recalled that there are a 
number of other Swahili words whose meanings could be extended to the 
concept of accessories: kingaja 'bracelet', vifuasi 'accessories' 
(accepted previously as the standard translation in the field of home 
economics), and ngaha 'traditional adornments'. The latter word is of 
Kinyaturu origin and refers to skins and body paint worn by performers at 
traditional drum and dance festivals ( ngoma ) . Delegates felt that the 
concept of accessories in the theatre was sufficiently different from that 
in the area of home economics that they rejected the previously adopted word 
vifuasi . They also felt that kingaja , as an item of female adornment only, 
was too narrow in meaning and would require too much semantic extension. 
Finally, the word ngaha was adopted since it refers to accessories. The 
fact that its use is in the context of traditional performing arts, rather 
than modern, is symbolic of the mixture of traditional and modern cultures 
which characterizes contemporary Tanzania. ^ 

BAKITA's Effectiveness 

BAKITA disseminates its decisions through its publications and weekly 
radio program. The pamphlets of official translations, Tafsiri San if u are 
sent to the relevant offices in the ministries, to schools and colleges, to 
the publishing houses and to the bookshops for sale to the general public. 
It is expected that those who are using Swahili in an official capacity, 
either in government or education, will refer to the lists for their 
vocabulary needs. However, in actual fact, BAKITA's terms are not widely 
used. Several evaluative studies show that the terms are used most among 
the educated, urban, wage-earning sector while indications are that the 
general public may not be aware of the existence of BAKITA (Ohly, 1982; 
Mdee, 1980). Specialists find that the terms lack clarity and do not exist 
in sufficient quantities to meet their needs. On the other hand, those who 
communicate with Tanzanians who have little or no formal schooling find the 
terms to be too technical, raising the complaint that contemporary Standard 
Swahili is too hard. Perhaps the most faithful users of the new vocabulary 
are those who participate in the creation of the lists of words themselves, 
particularly if their purpose in preparing the words was to enable them to 
write textbooks which would be accepted for use in the schools. 

While a certain amount of resistance to the new terms is to be 
expected, both from the specialist and from the layman, the form in which 
the new vocabulary is disseminated, as well as breakdowns in the channels of 
dissemination contribute to the problems which BAKITA encounters in 
implementing the use of their product. Simple lists of translational 
equivalents are not sufficient to enable most users to understamd precisely 


what concepts the terms label. Furthermore, since most of the lists are 
alphabetized according to the English equivalent, their use is limited to 
those; with knowledge of English. The monolingual Swahili dictionary is a 
general dictionary and does not include most of BAKITA's terms. 

Dissemination of the terms also encounters problems on the basic 
physical level. Although an office in a ministry may receive a copy of the 
official list of translations, there is no assurance that the list will 
reach the actual users. Most offices lack the machines and materials to 
reproduce the portions of the lists relevant to their activities as was 
originally expected. In other cases, the publications of BAKITA do not make 
it through the channels and thus never reach the hands of the intended 

Perhaps the most fundamental problem which limits BAKITA's 
effectiveness is the view of the vocabulary expansion process which 
implicitly informs the organization's activities. At BAKITA lexical 
elaboration is seen as the adding of words to the language. Despite the 
fact that the words to be standardized are often brought to BAKITA by the 
potential users, who are specialists in the subject area, BAKITA does not 
conduct a "needs analysis" to determine the characteristics of the users. 
At the same time, although subject specialists are involved in preparing 
lists of Swahili equivalents for English terms, BAKITA does not require 
these specialists to outline the concepts and their interrelationship so 
that related concepts will be labeled by terms which reflect this 
interrelationship. Furthermore, no distinction is made between the levels 
of technicality at which the same subject may be handled, a difference which 
might be reflected in the vocabulary to be created. As a result of this 
approach, concepts are assigned labels without attempting to preserve their 
interrelationship, thereby producing inconsistencies, while terms are 
adopted without considering the level of technicality at which a given 
audience conceptualizes, producing terms which are "too hard" for general 
use but which appear "fuzzy" to the specialist. 

One way to improve the success of BAKITA's efforts would be to adopt a 
model for vocabulary planning such as that proposed by Marshad for the 
development of Swahili technical terminology in Kenya (Marshad, 1984). 
According to this model, vocabulary expansion would proceed on two levels 
simultameously. Terms for the more technical levels would be adapted to the 
sound and spelling system of Swahili (following specified guidelines for 
nativization) from English or from an international standard. At the less 
technical levels, such as primary education, where little or no knowledge of 
English is assumed, terms would be coined from Swahili roots, following 
guidelines in order to insure that the relationship of concepts would be 
reflected through repetition of roots and affixes. While Marshad' s model 
could not be adopted wholesale in Tanzania, where, unlike Kenya, English is 
not used as the medium of instruction until secondary school level, it is 
offered here as an illustration of the type of refinement of the vocabulary 
expansion process that a model permits. 

Such a model assists planners in distinguishing the purpose for which 
their product is used, e.g., primary school, high school, adult education, 
etc., and thus lets them better gauge their intended audience. The model 
would provide a framework for the semantic and lexicological analysis of the 


relation of concepts to each other and to the terms which label them. The 
model also takes into consideration the needs of the society and the 
allocation of languages in the verbal repertoire (e.g., transitional 
educational bilingualism) . Once such a model is formulated, guidelines can 
be proposed for vocabulary expansion for each audience. As experience has 
shown in Israel, guidelines need to be formulated both for the general 
vocabulary expansion process and for the specific subject area and audience 
at hand (Fishman and Fellman, 1977). 

Until recently the lack of a clearly stated educational lamguage policy 
in Tanzania made the formulation of a model for vocabulary expansion 
impossible. That is, without knowing when or if Swahili would be 
implemented as the medium of secondary education, planners did not know who 
the users of their terms would be. Now that the policy has been clarified 
emd it is well known that Swahili will not be used as the medium of most 
post-primary education, BAKITA can focus its efforts on terms suitable for a 
primary school audience and for recent adult literates. Within a model of 
vocabulary expansion, this would involve determining, in conjunction with 
subject area specialists who deal with these audiences, the level of 
technicality as well as the concepts to be taught at this level. Guidelines 
for the preparation of vocabulary could then be written with the needs of 
these audiences in mind. Closer cooperation with individuals in education, 
publishing and the media who would be in a position to disseminate the terms 
would help to insure their exposure. The publication of monolingual subject 
(classified) glossaries might also be an effective means of disseminating 
the new vocabulary. 

Vocabulary expansion efforts in Tanzania, and in the Swahili speaking 
world in general, would also benefit greatly from the documentation of the 
terminological efforts to date that would be provided by a term beink such as 
that which is in the pilot stages at the Institute of Kiswahili Research 
( Tumbo-Masabo , 1983). At present, much of the rationale used for adopting 
various terms resides in the memories of the employees of BAKITA. A term 
bank would allow for the recording of terms as well as the guidelines which 
motivated their adoption. At the same time, the term bank, if organized 
according to international classification principles as suggested by 
Tumbo-Masabo, would allow planners to better delimit the fields for which 
terms are being created. If the bank were further organized according to 
concepts within each field, with terms filed according to the concept each 
labels, competing terms could easily be compared and lexical gaps 
identified. Eventually it is hoped that BAKITA and/or the Institute of 
Kiswahili Research would use a micro-computer for storage and manipulation 
of the term bank. 

It is unreasonable, however, to expect BAKITA to implement these 
suggestions on its own. The low operating budget and the low level of 
expertise of its academic employees seriously impede the organization's 
effectiveness. Without adequate support, both in terms of financing and 
training, BAKITA cannot command the authority with which it is legally 
invested. It remains to be seen whether, under the new educational language 
policy, BAKITA will be able to redirect vocabulary expansion in Swahili to 
serve the needs of those users identified by the policy. Tanzania is in the 
leading position in the modernization of Swahili, and there is much 
enthusiasm and experience at BAKITA, the Institute of Kiswahili Research, 


the Department of Kiswahili at the University of Dar es Salaam, as well as 
among members of the general public for the expanded use of the language. 
The realization of Swahili's full potential depends to a large extent upon 

firm policy backing. 


'The research on which this paper is based was conducted in Dar es 
Salaam while I was a Research Associate in the Department of Kiswahili from 
August, 1983 to June, 1984. My sincere thanks are due to the employees of 
BAKITA who graciously welcomed me in their offices amd at their meetings, as 
well as to the Institute for Kiswahili Research. My appreciation also 
extends to Z. M. Mochiwa from whose comments I have benefited immensely. 
The research was funded by a Fulbright Grant for Graduate Study. 

^ The label Swahili, however, masks the dialect variation which exists 
in the language. A number of dialects of Swahili are spoken as native 
languages in settlements along the coast and on the islands of Zanzibar and 
Pemba. Monolingual speakers of Swahili are also found in urbam areas of the 
mainland. The majority of Swahili speakers are bilingual speakers of the 

^Tanganyika was colonized by Germany from the late nineteenth century 
until World War I. Britain then controlled the colony until its 
Independence in 1960. Zanzibar remained an independent sultanate until the 
revolution in 1964. Later that year Tanganyika and Zanzibar merged to form 
the United Republic of Tanzania. In 1975 the separate governments of 
Zanzibar and the mainland were united under the supremacy of the political 
party, Chama Cha Mapinduzi . 

''1967 was a landmark year in the establishment of Tanzania's socialist 
orientation towards political, economic and social development with the 
publication by President Julius Nyerere of the Arusha Declaration. 

5 'Akizungumza lugha ya Kiingereza nchini, Mwalimu alisema kuwa 
Kiingereza ndicho Kiswahili cha dunia na kwa sababu hiyo hakina budi 
kufundishwa nchini na kupewa uzito unaostahili. "Kiingereza ni Kisweihili 
cha Dunia. Ni makosa kukiachia Kiingereza kikafa. Kukiachia ni ujinga siyo 
uzalendo," alisema. Mwalimu aliwataka Watanzania kung'ang'ania Kiingereza 
na Kiswahili na kuongeza kwamba kitakuwa ndiyo lugha ya kufundishia katika 
shule za Sekondari na Vyuo vya juu kwa sababu kikiachwa kama somo la kawaida 
kinaweza kufa. ' Mzalendo . October 28, 1984. 

* For example, rara 'ballad' which comes from Yoruba and was adopted at 
the November, 1983 meeting of the language stemdardization committee at the 
suggestion of a Yoruba-speaking Nigerian graduate student in the Department 
of Swahili at the University of Dar es Salaam who was observing the 

''The November, 1983 meeting of the standardization committee had 
approximately 35-40 participants including delegates, employees of BAKITA 
and guests. The August, 1982 meeting had 20 delegates alone. 


8 Thf'se observations were made at the November, 1983 meeting of the 
standardization committee which was held in Iringa, Tanzania. 


ABDULAZIZ, M. 1972a. Tanzania national language policy and the rise of 
Swahili political culture. In Socialism in Tanzania: an 
interdisciplinary reader, ed. by Lionel Cliffe and John S. Saul. 
Nairobi: East African Publishing House. 

1972b. Triglossia and Swahili-English bilingualism in 

Tanzania." Language in Society, I. 

BARAZA LA KISWAHILI LA TAIFA. 1969-1981. Taarifa za mwaka. Dar es Salaam: 
Government Printing Office. 

1974, 1976, 1978, 1980. Tafsiri sanifu. 

Dar es Salaam: National Publishing Company. 

n.d. Kamati ya kusanifu lugha. 

1983. Jifunze Kiswahili uwafunze wengine. 

Dar es Salaam: Government Printer. 
BRAUNER, S., C. Kapinga and K. Legere. 1978. Swahili and local languages 

in Tanzania: a sociolinguistic study." KiswEihili Vol. 48/2. 
DUNN, A. 1984. Tanzania's Swahili language policy: a critical view of 

implementation." Paper presented at the Univeristy of Illinois African 

Studies Roundtable Meeting on "Language, Literacy and Communication in 

Africa." November 8-9, 1984. 
FELLMAN, J. and J. Fishman 1977. Language planning in Israel: solving 

terminological problems. in J. Rubin, et al. Language planning 

processes. The Hague: Mouton. 
HEINE, B. 1976. Knowledge and use of a second language in Musoma Region- 

a quantitative survey. Kiswahili. Vol. 46/1. 

dictionary. Nairobi: Oxford University Press. 
IRIRA, S. 1982. Kutohoa maneno. Paper presented at the 16th meeting of 

the Committee on Language Standardization, Dar es Salaam, August 9-18, 

1983. Kusanifu lugha ya Kiswahili. Paper presented at a 

seminar for language and cultural officers of Coast region, April 7-8, 

KISHE, A.J. 1982. Taarifa ya utafiti juu ya maneno yanayoandikwa tofauti na 

yanavyotamkwa. BAKITA, Kamati ya Sarufi na Ithibati. 
KLOSS, H. 1969. Research possibilities on group bilingualism. Quebec: 

International Center for Research on Bilingualism. 
MACWILLIAM, A. 1983. Some thoughts on translation of scientific terminology 

in Swahili. Dar es Salaam: Institute of Kiswahili Research. 
MARSHAD, H. 1984. An approach to code elaboration and its application to 

swahili. Urbana, IL: unpublished PhD dissertation. 
MDEE, J. 1980. The degree of acceptability of new Swahili terms: speakers' 

response analysis. MA thesis. University of Dar es Salaam. 
1983. The policy of adapting loan words in Kiswahili as conceived 

of by BAKITA: a critique. Multilingual, 2-2. 
O'BARR, W. 1971. Multilingualism in a rural Tanzanian village. 

Anthropological linguistics, 11. 
OHLY, R. 1982. Report on the state of modern Swahili in urban Bukoba (May 

1978). Kiswahili, 49/2. 


POLOME', E. and C.P. Hill. 1980 Language in Tanzania. International African 

Institute; Oxford University Press. 
SAMBAZI, M. 1984. Shughuli za tafsiri katika Baraza la Kiswahili. Paper 

presented at the Translator's Conference, University of Dar es Salaam, 

Feb. 2, 1984. 
TAASISI YA UCHUNGUZI WA KISWAHILI. 1980. Kamusi ya Kiswahili sanifu. 

Dar es Salaam: Oxford University Press. 
TUMBO-MASABO, Z. 1983. Jinsi ya kuanzisha benki ya 

ist ilahi/msamiat i . Paper presented at the International 

Standardization of Terminology in Swahili, Dar es Salaam, 

WHITELEY, W. 1957. The work of the East African Swahili 

1930-1957. Kongo-Overzee, Vol. XXIII. 
1969. SweJiili; the rise of a national language 


data ya 

Meeting on 



London : 

studies in the Linguistic Sciences 
Volume 15, Number 1, Spring 1985 


Hans Henrich Hock 

The feasibility of syntactic reconstruction has been doubted in a number of recent 
publications In this paper 1 show that the theoretical arguments In favor of this negative 
view are dubious The paper concludes with an empirical demonstration that syntactic 
reconstruction can and does yield acceptable results, provided that it is founded on the 
cumulative evidence of 'archaisms'. The data for this demonstration come from Romance, 
the verification of the approach, from Latin. 

I: In recent years a number of linguists have doubted the feasibility of syntactic recon- 
struction; cf. eg Campbell & Mithun l980,Jeffers 1976, Lightfoot 1979, 1980, et alibi, Mithun 
& Campbell 1982. Winter 1984 The major reasons given for these doubts are: (a) The com- 
parative method works well for phonology because of the regularity and built-in directionality of 
sound change, but fails to work for syntax because syntactic change is not regular and has no built-in 
directionality Moreover, while phonological (and morphological) reconstruction can work with 
cognate WORDS , there are no cognate SENTENCES For excepting frozen expressions, sentences are not 
Inherited, but created anew by each speaker. The basis for this creation is a system of internalized 
rules, hypothesized to account for the surface structures of an earlier generation And In the process 
of hypothesizing such rules, reinterpretatlon, an inherently unpredictable and non-directional 
process, plays a major role, (b) Typology, which has been Invoked by scholars as diverse as 
Lehmann ( 1974, 1982) and Friedrich ( 1975, 1976, 1977), does not provide a reliable guide, 
since there is no need to expect systems of proto- languages to be typological ly any more consistent 
than those of attested languages Moreover, as 1 have shown elsewhere ( 1985), the expectation of 
typological consistency In proto- languages may lead different scholars to diometricolly opposed 
conclusions, based on essentially the same data 

The second of these two arguments is well taken and will be given further support in this paper 
The first, however, leaves much to be desired For as I will demonstrate, the claimed fundamental 
difference between phonological and syntactic reconstruction is spurious In addition, the general 
conclusion, that syntactic reconstruction is impossible, will be shown to be unacceptable 

2: Indo-Europeanists have for a long time, since et least Delbriick ( 1893-1900), proceeded 
under the assumption that syntactic reconstruction is possible However, until recently this 
assumption did not come under attack, and no theoretical or methodological defense of syntactic 
reconstruction was considered necessary Without such a defense, however, the claim that syntactic 
reconstruction is impossible cannot be meaningfully refuted And the fact that Lehmann { 1974, cf. 
also 1982) and Friedrich (1975, 1976, 1977) have advocated diametricQlly opposed recon- 
structions for PIE word order may be taken to confirm the view that syntactic reconstruction simply 
is not feasible 


3: Here as elsewhere, of course, we must be careful not to interpret scttolarly disagreement as 
an Indictment of an entire scientific approacti And an absence of theoretical or methodological 
statements does not necessarily signal a lack of implicit theory or methodology In fact, the 
divergence between Lehmann and Friedrich has elicited in Watkins 1 976 a more explicit formulation 
of the traditional approach to syntactic reconstruction, as being based on the comparison of syntactic 
PATTERNS in related languages, with preference given not to the regular patterns, but to shared 
ARCHAIC ABERRANCIES For it is the latter which are more likely to preserve an earlier state 
Moreover, a reconstruction which can account for the synchronlcally regular patterns as innovations 
and for the aberrant ones as archaic relics of earlier, different regularities provides a more 
satisfactory explanation of the facts than one which is restricted to the synchronic regularities 
Reconstruction thus boils down to proposing a hypothesis which gives the best dynamic account of the 
various patterns encountered in related languages. ' 

4: The very usefulness of the notion archaism' , however, Is guesttoned by Friedrich He rejects 
the notion mainly implicitly, by basing much of his reconstruction on the most prevalent patterns 
found in the various Indo-European languages But in his discussion of the applicability of 
Kurylowlc's Fourth Law to syntactic reconstruction ( 1 976: 2 1 3), he explicitly rejects the concept 
His argument can be paraphrased as follows; How do we know that a given synchronic aberrancy is an 
archaism, rather than, say, an incipient or abortive innovation? it is circular to argue that a given 
structure X is archaic because it agrees with a reconstructed pattern Y, if that reconstruction is based 
on the claim that X is archaic. 

We have here, then, a very serious argument against the traditional approach, an argument 
which has been adopted also by some of those who have claimed that syntactic reconstruction is 
Impossible (cf. e.g. Lightfoot). 

5: Problems arise also in respect to the notion 'pattern': Some linguists, notably Lehmann ( but 
cf e.g. also much of Friedrich's work), have characterized their approach as the reconstruction of 
RULES It is not always clear whether the term 'rule' here is simply a trendy synonym for 'pattern' 
or whether it Is understood to mean something distinct, namely the mechanism' which generates' or 
'accounts for' a given pattern within a synchronic grammar. If in fact the latter is meant, then it 
would certainly be justified to argue that such an approach is not feasible, since rules are not handed 
down from one generation to another, but only utterances from which such rules are abstracted' 

6; Superficially, the same argument might apply also to reconstructions based on syntactic 
PATTERNS, since these, too, are 'abstracted' frcm the utterances of earlier generations Just like 
sentences and rules, they are not inherited, but crrated anew by each speaker The absence of cognate 
structures, then, makes reconstruction Impossible for syntax -- unlike phonology and morphology, 
where cognate words furnish a basis for reconstruction. Moreover, it might be claimed, the 
distinction between patterns and rules is meaningless. 

This argument, however, Is open to doubt. First, as Hale's (1973) famous Maori example 
shows, a PATTERN of consonant/0 alternation can give rise to very different RULE systems. (Cf also 
Ohala's ( 1974) a_/0 alternations in Hindi ) There is thus a difference between patterns and rules; 
and patterns are more basic or primary than rules. Moreover, unlike sentences and rules, patterns 
are recognized, not 'created'. 

Secondly, the distinction between phonology and morphology on one hand and syntax on the other 
is not as great as claimed. Sounds and morphemes must just as much be abstracted' from the 


utterances of earlier generations as sgntactic patterns True, the former can be abstracted from 
individual words, the latter cannot; and words in some sense are more concrete and learnahle than 
sentences But the distinction is only one of degree For just like sentences, words do not come with a 
ready-made analysis; but their phonological and morphological patterning must be recognized by 
the child Moreover, both In syntax and In phonology and morphology, It Is only through the 
recognition of such patterns that the child can successfully develop an internalized grammatical rule 
system And just as the child does not need to have heard all sentences to recognize the major 
syntactic patterns and to account for them In terms of rules, so It Is not necessary to have heard all 
the words of the language to recognize the relevant phonological and morphological patterns and to 
hypothesize rule systems that account for them Finally, just as reinterpretation and creative 
extension are an ever-present possibility in syntax, so they are in the lexicon, even if to a lesser 

7: Historical and comparative linguistic evidence and considerations support the view that the 
distinction between syntax on one hand and phonology and morphology on the other is not as great as 

First of all, Labov's work has shown that linguistic change takes place primarily in the peer 
group of adolescents or adults, not in children's language acquisition (Cf also Pybee and Slobin 
1982 ) Arguments which focus on alleged differences between syntax and phonology/morphology in 
first language acquisition therefore are not particularly germane to historical and comparative 

Secondly, reinterpretation Is found not only In syntax or In morphology (cf ( I ) and (2)), but 
also in phonology (cf. (3)) (In fact, some linguists would go so fares to claim that all sound change 
results from phonological reinterpretation; cf. Andersen's ( 1973) notion of abductive change ) 

( 1 ) Syntactic reinterpretation; 

(a) Skt. aranyevrkah Ui balakah tatra nagacchati 

'forest'wolves' quote 'boy' 'there' 'not' goes' 
'( Think Ing) "There are wolves In the forest" , the boy does not go there' 
-» 'Because there are wolves in the forest, the boy does not go there' 

Hence, with reinterpretation of quotative Hi as causal marker: 

(b) aranye vrkah tti avayah tatra gatvS ksfyante 

'sheep" there"havinggone"perish' 
'Because there are wolves in the forest, the sheep, having gone there, perish' 

(2) Morphological reinterpretation: 

(a) Early NEngI pease (porridoe hot) -- a mass noun 

(b) NEngI. pea ( so ) : pea-s (pi.) -- with reinterpretation of 
[-z] as plural marker 

(3) Phonological reinterpretation 

Du (*)luft > luxt air', Rom * direktate - > dereptate justice, with 
acoustically-based reinterpretation of [ + grave, - velar) as [ + grave, ♦ velar] or 
vice versa. 






*t)ar-br {'>) 



•hwar-bT (7) 



•t)ar-in(ne) {->) 



*hwar-in(ne) (?) 









Moreover, just as in syntax we can trace the fate of patterns but not of individual sentences, so 
In morphology it is often only possible to trace the fate of a given derivational or inflectional pattern 
but not necessarily of words. Thus, the correlation between English and German words of the type 
thereby : dabei . whereby : wobei might suggest reconstructions as in (4). However, the fact that both 
German and English have words which are not matched (5a), or only incompletely matched (5b) by 
the other language should suggest that it is the pattern which is inherited and not any particular 
words. 2 This suspicion is confirmed when we look at the history of these structures Words of the 
type thereby are found In English and German as early as the 9th century; the whereby type, on the 
other hand, appears to be first and guite sparingly found in late lOth/early 1 Ith-century German 
(Notker, four attested forms) and 12th-century English (Hatton Gospels, one form). In both 
languages, the bulk of the forms is of later manufacture 



8: Also in respect to regularity and built-in directionality, the difference between syntactic and 
phonological or morphological change is not as great as claimed. True, sound change is 
overwhelmingly regular (cf. Hock 1976b for a recent defense of that claim); but as noted for 
instance by Miranda (comment on Jeffers 1976), that does not mean that correspondences will be 
regular. Other changes, such as analogy or semantically induced tabooistic distortion (as in Engl 
shoot for the well-known expletive), provide an ever-present source for irregularity In addition, 
some sound changes, notably metathesis and dissimilation, are notoriously irregular (For conditions 
under which they may be regular, cf. Hock In Press (a. b).) -- Analogical change typically is not 
regular, but rule-governed changes may be as regular as sound change, cf eg the German 
generalization of prepausal devoicing to word-final or syllable-final environment (Hock 1976a ) 
-- Syntax, Deing eminently rule-governed, Dy definition commonly operates with very sweeping, 
regular changes, such as the English generelization of object to subject promotion in ( 6) Moreover, 
the fact that analogy and syntactic change are not as regular as sound change may actually be a help in 
reconstruction, since relics of an earlier stage may be preserved Compare the English irregular 
plurals pence and dire which suggest that the regular pattern in dens or files is an innovation (cf 
Hxk 1 976b). In the area of regular sound change, however, such archaisms by definition are absent 

(6) (a)OEng icseohine heisseon (Acc.DO -» Norn S) 

ic help him himisholpen (Dat. "DO" remains) 

icgiefehim himisgiefen (Dat. 10 remains) 

(b)NEng I see him he is seen (D0-> S) 

I help him he is helped (ordinary DO -> 5) 

I give him a book he is given a book (10 -» 5) 

The issue of built-in directionality likewise does not provide a clear-cut distinction between 
syntactic and other change First of all, Vincent (1980) correctly notes that the gremmatic- 
alization' of lexical items, as in ( 7) is generally irreversible ( if we except rare opposite develop- 
ments like Engl, to up one's income ).^ Similarly, the common process of placing sentence clitics into 


clause-second position may pnxluce structures like ( 8) in original SOV languages, with AUX or finite 
verb In second position and the nonftnlte participle clausfi-flnaUcf Sififlle 1977, Hock 1982); but 
I am not awarB of any process wtiicti would bring about such structures in original SVO languages 
True, directional developments like these probably constitute a small minority in syntax But 
morphological change fares no better What seems to be less commonly realized is that also most 
sound change has no built-in directionality To be sure, some developments, such as palatalization 
seem to be irreversible as sound changes, cf (9) But the majority of processes, including assim- 
ilation, are far from having a unique directionality, cf. ( 10). (See also Miranda 1978.) 

(7) Ital. casa. Span casa 'house' Fr. che2{ prep.) at the house of, with' 

(8) Oltal. io fui al pie dun colle qiunto 'I had come to the foot of a hill' 

AUX pple. 

OEng her wees Crist ahanqen 'Here Christ was hanged ( = crucified)' 

AUX pple 

(9) Palatalization: k > c etc. (butnoc > ketc.) 

( 10) Assimilation: tm > pm ( Swiss opma = NHO atmen ) 

pp (Skt. atman - : hIAr dial appan -) 

nm (Kor patmada > panmada ) 

mm (Aeol. Gk kat(a)moros > kammoros ) 

tn ( late Anc Gk dial Patmos > Patnos ) 

tt (Skt. atman -: MIAr dial attan-) 


9: The preceding discussion shows that the claimed fundamental distinction between syntax on 
one hand and morphology and phonology on the other does not hold In principle, therefore, 
reconstruction should be as possible -- or, heaven forfend. Impossible --in syntax as It Is In 
phonology and morphology. This is not to say that reconstruction will always be possible or uncon- 
troversial After all, even in Indo-European phonology, the best and longest researched area of 
comparative linguistics, controversies continue on such topics as the exact number and phonetic 
identity of the 'velars' and of the so-called laryngeals, or the phonetic, phonological, and typological 
characteristics of the entire obstruent system. 

1 0: In the remainder of this paper I will demonstrate that syntactic reconstruction is possible 
not only IN PRINCIPLE but IN PRACTICE Given the wide-spread scepticism about the feasibility of 
syntactic reconstruction, such a demonstration would certainly seem desirable, but as far as I can 
tell, it has not so far been undertaken. In addition, I will show that the traditional approach with its 
emphasis on aberrant, archaic patterns can be applied without circularity and that It yields more 
satisfactory results than an approach which focuses on the synchronically most regular patterns 
Finally, I will show that attempts to reconstruct a typologlcally consistent proto- language may lead 
to wrong results The data for this demonstration come from Romance word order; the verification of 
the method, from Latin.'^ 

1 1 There is virtually universal agreement that in Latin, the ( near- )ancestor of the Romance 
languages, the pragmatically unmarked order of the ma)or constituents was SOV And as In other SOV 
languages, the order of auxiliary (AUX) end non-finite main verb (mv) most commonly was MV + 
AUX (cf Hock 1982) Beside SOV, various other orders were possible, many of which can be 
explained as due to stylistic NP movement processes In addition, a marked verb- Initial pattern 


existed, in which the fronted verb signaled fxus on the action, stage- setting, etc. Compare the 
examples in ( 11 ). 

(II) (a) S(0)V* hi omnes lingua, Instilutls. legibus inter se differunt 

'all of these differed from each other in speech, political institutions, 
and laws' (Bell. Gall. 1.1.2) 

( b) MV+ AUX cuius pater .. a senatu populi Romani amicus appelatus erat 

'whose father had been called friend by the Roman senate and people' 

( c) *V erant omnino itinera duo 

'(there) were only two roads' (lb. 1.6. l) 

(d) Scrambl. ea res est Helvetiis per indicium enuntiata 


'this matter was made known to the Helvetians through spies' 
(ib.l 4.1) 

The question arises as to whether reconstruction on the basis of modern Romance evidence can 
approximate this state of affairs. 

11: The synchronically most productive and unmarked pattern of the Romance languages clearly 
is SVO, with the order AUX + MV; cf. e.g. ( 12). If we were to use this evidence as the basis for 
reconstruction, then we would fail to arrive at the unmarked SOV and MV + AUX of Latin. This 
approach therefore is Inadequate, and we must look to the traditional , archaism -based approach for an 













pom me 












1 2: In order to be acceptable, however, this approach requires a non-circular determination of 
what constitutes an archaism , based on independently established criteria. 

One such criterion Is the traditional wisdom of historical linguistics, codified In Kurytowlczs 
( 1 947) 'Fourth Law of Analogy' , that if in non-phonetic change OLDER forms continue to exist next 
to the innovated forms, they are LIMITED TO MARGINAL FUNCTIONS. Moreover, observation of attested 
linguistic histories shows that certain types of language use tend to best preserve such marginal 
forms. These conservative texts include legal documents, and traditional literary forms, such as epic 
poetry, popular ballads, and folk narratives. 

At the same time, not all deviant patterns are necessarily archaic Some may be mistakes, or 
Intentional deviations (especially in non- traditional poetry), or the harbingers of a coming 
innovation. Such deviations however would not be be limited to conservative texts. Moreover, if such 


deviations are synchronically felt to be 'old-fashioned' and if within observable history their use can 
be seen to be decreasing, it is more likely that we are dealing with archaisms than with, say, 
incipient innovations. 

1 3: The case for considering a particular pattern an archaism becomes even stronger if we can 
find CUMULATIVE evidence in related languages, such that they show similar deviations, but in very 
different contexts, at least some of which are typical for archaisms. In that case, common innovation 
would be highly unlikely, and the only satisfactory historical explanation will be the hypothesis 
that we are dealing with common inheritance. 

1 4: In addition to the productive pattern of ( 1 2), the Romance languages also have marginal 
patterns, i.e. possible archaisms. One of these is pan-Romance, although outside imperatival 
structures it appears to be on the retreat, especially in French where only a few frozen expressions 
sur- Vive. This is the verb-initial pattern in ( 13). Disregarding the still-productive imperative 
structures, this pattern serves to focus on the verbal action and as a stage-setting device. Since this 
pattern is marginal It may possibly be an archaism, of which the predominant SVO would be a 
secondary replacement. However, the fact that all the Romance languages agree on the marked 
functions of this construction suggests that its marginality, including the markedness of its 
functions, is inherited from the proto- language, i.e. that in the proto- language verb-initial order 
coexisted and contrasted with the ancestor of the modem unmarked SVO. 

This impression is reinforced by the fact that a similar distinction between verb-initial and 
other sentence structures is made in the case of verb + clitic pronoun sequences, at least in some of 
the languages; cf. ( 14). And the fact that In French these structures are limited to the only 
productive verb-initial type, the imperative, and that Italian and Spanish have a strong tendency 
toward a similar limitation suggests that the contrast between verb-initial and other sentence 
structures is an archaism, not an innovation. 

(13) (a) 


vaya con Dtos 

■go with God" 



'look for the woman' 


lasciate ogni speranza 

'abandon all hope' 



fueenestacludadunrey . 

.. 'there was in that city a king' 


suivent lesnoms... 

'(now) follow the names ...') 


disse il professore ... 

'said the professor.,.' 

(14) (a) 


mostrad me 

'show me' (impve.) 

ven^mos lo 

'wewill ^thaf (non-impve.) 



'tell it' (impve.) 



'show me' (impve.) 

disse gli ladama... 

'said the lady to him ...'(non-impve.) 

vs. (b) 


(yo) teamo 

"I love gou' 


je t'aime 

" " 


(io) ti amo 

15: There is evidence also for marginal (S)OV, verb-final, or MV ♦ AUX patterns. 
However, unlike the verb- initial type, these latter patterns are found in very different contexts in 
the various Romance languages. And these contexts Include traditional (popular) poetry. 


Evidence forOV is found in the so-called ^-optative of Romanian, as in ( 15). It is also found 
in frozen, 'idiomatic' expressions like ( 16a). Contrast what would be the productive pattern in 
( 16b) with VO AND article. 

Verb-final patterns can be observed in Romanian andGalician folk poetry of the last century, as 
In (17). 

A combination of verb-final and MV + AUX is frequently encountered in Sicilian and Sardinian 
dialects, cf. (18). 

Mv + AUX is found in 'frozen' form in the non-Balkan Romance future That forms like ( 19a) 
are in fact untverbations of infinitive + 'have' is shown by the evidence of Portuguese and Sicilian, 
which permit a clitic pronoun to inter/ene; cf ( 19b). Romanian uses a different auxiliary, namely 
the verb 'want', which may precede or follow its main verb;cf. (20a) And the same option exists for 
the aux 1 1 tary of the condl t tonal ; cf . ( 20b ). 

( 1 5) Rom. dereptate s(a) ave^i spre tot omu-lu 


that you may have justice for every man' 

(16) (a) Fr. sansmotdire 'without saying a word' 


(b) Fr. sans dire UQ mot 


(17) Rom. cu nevesta me lubesc / dupa fata me topesc 

V V 

"I am in love with the woman, a pine away for the girf 

(18) Sic. lapicilidda vatiattae? 'has the girl been baptized?' 


(19) (a) Span. amare = amar + he 'I will love' 

Inf. have 
(b) Port. faze-lo-hei 'I will do it' 

(20) (a) Rom. voT cinta / ctntavoT 'I will sing' 

(b) Rom. a^T ctntW ctntd a§T 'I may, would sing' 

16: What is important about these marginal patterns is that the Romance languages diverge 
greatly, showing no cross- the- board agreement either on the specific patterns or on the contexts in 
which they are used At the same time, the patterns all point toward the same word order, namely 
SOV, with verb in final position, andMV + AUX 

If we reconstruct SOV as one of the orders of the proto- language, then we are able to account for 
these divergent patterns as archaisms, whose survival in a given language and a particular context 
con be a matter of accident. On the other hand, if we do not reconstruct SOV, then we wilt have to 
claim that for some strange and unexplained reason, all of these languages independently underwent 
Innovations which, though taking place In very different contexts, 'happened' to converge In the 


direction of SOV. Clearly, of these two alternatives, the former provides the better explanation and is 
to be preferred. 

1 7: What remains to be settled is whether SOV should be reconstructed as an alternative to the 
predominant 5V0 of modern Romance, or In Its stead. Evidence for original Identity between SOV and 
SVO may be found in the fact that no special, cross-Romance function seems to be attached to the SOV 
patterns and that they are just as much in contrast with the specially marked verb- initial pattern as 
Is the synchronically predominant SVO. Also the clitic patterns in ( 1 4) argue for an original contrast 
between TWO patterns (verb- initial and 'other'), not three (verb- initial, SVO, and SOV)^ in view 
of these arguments, It seems preferable to reconstruct SOV as the original unmarked order and to 
consider SVO an innovation 

A possible counterargument might be that this would entail the dubious assumption that all of 
the Romance languages independently innovated by changing SVO to SOV However, given that (most 
of) the neighboring non-Romance languages likewise have SVO and that Romance shares other 
features with these languages (such as a contrasting pair of definite and indefinite articles), the 
change of SOV to SVO may be considered a common, areal innovation 

18: The reconstructions we are left with, then, ere verb- initial and SOV And the fact that 
these are the two major contrasting patterns of Latin shows that the traditional approach to syntactic 
reconstruction, with Its emphasis on the notion 'archaism', can yield results which closely 
approximate the (near-)ancestral language Syntactic reconstruction, thus, is possible True, the 
evidence examined would not enable us to reconstruct the scrambling exemplified in ( 1 Id), or to 
determine the precise route by which SOV changed to SVO But as noted earlier, areas of uncertainty 
exist even in phonological reconstruction. 

19: Let me conclude with a brief glance on the relevance, or lack thereof, of typology for 
syntactic reconstruction. If we were to follow the typologistic approach of scholars like Lehmann and 
Friedrlch, we would run Into the difficulty that all the Romance languages have postnomlnal relative 
clauses, while 'typological ly pure SOV languages employ different strategies To judge by these two 
scholars' approaches to Indo-European reconstruction, Lehmann would therefore conclude that the 
evidence of the Romance languages notwithstanding, the ancestral language, being SOV, must be 
reconstructed as having pronominal relative structures, while Friedrich would consider the 
postnomlnal relative clauses evidence against an ancestral SOV In fact, however, the relevant 
chronological layer of Latin had SOV as its unmarked major constituent order, as well as post- 
nominal relative clauses. And given the Romance evidence, this is precisely what would be postulated 
by a reconstruction which does not try to Impose on the ancestral language what Watkins has called 
the straight-jacket' of typology. 


*I am grateful to my colleagues Mario Saltarelli and Dieter Wanner who have commented on an 
earlier draft of this paper Needless to state, the responsibility for any errors or omissions rests 
with me 

'The argument that it is syntactic PATTERNS, rather then individual sentences, which form the 
basis for syntactic reconstruction is found also in Hall 1968,Gulstad 1974, and Costello 1983 But 


these publications do not address the importance of archaisms and thus do not offer any means for 
evaluating competing reconstructions Eckert ( 1978) comes closest to Watkins, by emphasizing 
fixed, archaic collxations as the basis for syntactic reconstruction However, he does not elaborate 
on whether and how reconstruction can go beyond the individual ancestors of these collocations 

^ This presupposes, of course, that the univerbalion of earlier * bar bT etc to OE fteerbiq etc , 
although attested only from the 9th century, began as a variable process in the ancestral language 
Otherwise, the development would be even more complex, although the basic argument would remain 

•^In this assessment, Vincent was anticipated by Miranda 1978 However, a number of the 
processes which Miranda cites as unidirectional are of doubtful cogency, including Givon's ( 1971 ) 
claims regarding Romance clitic ordering. ( For these claims see also below.) 

^Many of the Romance data are drawn from Richter 1903 and Rohlfs 1954 

^The fact that In ( Mb), the clitic object pronouns precede, rather than follow the verb might 
be taken as a further archaism of earlier OV order; cf. Givon 1971. This argument, however, is not 
cogent; for the preverbal position of object clitics can be accounted for by a different scenario; cf. 
Wanner 1 985 and Pearce 1 984. 


ANDERSEN, Henntng. 1973. Abductlve and deductive change. Language 49.765-93. 

Bybee , Joan L. , and Dan I. SL06IN. 1 982. Why small children cannot change language on their own: 

Suggestions from the English past tense. Papers from the Fifth International Conference on 

Historical Linguistics, ed. by A. Ahlqvlst, 29-37. (Amsterdam Studies In the History and 

Theory of Linguistic Science, Series IV, v. 29.) 
Campbell . Lyie, and Marianne Mithun 1 980. The priorities and pitfalls of syntactic reconstruction. 

Folia Linguistica Historica 1 1 9-40. 
Cos TELLO, John R. 1983. Syntactic change and syntactic reconstruction: a tagmemic approach. 

Arlington, TX: Summer Institute of Linguistics. , 

DELBRUCK , Bertold. 1893-1900. Vergleichende Syntax der indogermanischen Sprachen (=v 3-5 

of K. Brugmann, GrundriC der vergleichenden Grammatik der indogermanischen Sprachen.) 

StraDburg: TrUbner. 
ECKERT, Reiner. 1978. Syntagmatik und Rekonslruktion. Proceedings of the Twelfth International 

Congress of Linguists, ed. by W. U. Dressier et al., 477-80. Innsbruck: Innsbrucker Beitrage 

zur Sprachwissenschaft 
Friedrich, Paul. 1975. Proto-Indo-European syntax. (Journal of Indo-European Studies, 

Monograph 1.) Butte, Montana. 

. 1976. AdHrek. Journalof Indo-European Studies 4.207-20. 

. 1977 The devil's case: PIE as type II Linguisticstudiesoffened to Joseph Greenberg, ed. 

byAJulIland, 3.463. Saratoga, CA: AnmaLlbri. 
6Iv6n, Talmy 1971 Historical syntax and synchronic morphology: an archaeologist's field trip. 

Papers from the 7th Regional Meeting of the Chicago Linguistic Society, 394-415. Chicago: 

Linguistic Society 
QULSTAD, Daniel E. 1 974. Reconstruction in syntax. Historical linguistics, ed by J M Anderson & 

C. Jones, 1.117-58 (Proceedings of the First International Conference on Historical 

Linguistica.) Amaterdom: North Holland. 


Hale, Kenneth 1973. Deep-surface canonical disparities in relation to analogy and change: an 

Australian example Current trends in linouistics 1 1 .40 1 -58, ed bu T. S Sebeok. The Hague: 

Hall, Robert A , Jr 1968. Comparative reconstruction in Romance syntax Acta Linguistica 

Haffniensia 1 181-8 
Hock , Hans Henrich 1976(8) Current trends in historical linguistics. Lektos 2:1,25-48. 
1976(b) Review article on Anttila ( 1972) Introduction to historical and comparative 

linguistics Language 52 202-20 
1982 AUX-cliticization as a motivation for word order change Studies in the Linguistic 

Sciences 12191-101. 
1985. Review of Lehmann & Malkiel, eds ( 1982) Perspectives on historical linguistics 

Language 61 187-93 
in Press (a). Regular contact dissimilation. A Festschrift for Henry Hoenlgswald, ed by G 

Cardona and N. Zide Berlin et alibi; Mouton 

In Press (b). Regular metathesis Linguistics 234 ( 1985) 

JEFFERS, Robert J. 1976. Syntactic change and syntactic reconstruction Current progress In 

historical linguistics, ed by William M. Christie, 1-10. Amsterdam: North Holland 
KURYtOWiCZ, Jerzy 1947. La nature des procesditsanalogiques. Acta Linguistica 5 17-34 
LEHHAMN, Winfred P. 1974. Proto-Indo-European syntax. Austin: University of Texas Press 
1982 Introduction: Diachronic linguistics Perspectives on historical linguistics, ed by 

W P Lehmann and Y. Malklel, 1-16. (Amsterdam Studies In the Theory and History of 

Linguistic Science, Series 4, V. 24 ) Amsterdam: Benjamins. 
LIGHTFOOT, David 1979 Principles of diachronic syntax. Cambridge: University Press 
1980 On reconstructing a proto-syntax Linguistic reconstruction and Indo-European 

syntax: Proceedings of the Colloquium of the Indogermanlsche Oesellschaff ... 1 979, ed by P 

Ramat et al , 27-45. (Amsterdam Studies In the Theory and History of Linguistic Science, 

Series 4, V. 19) Amsterdam: Benjamins. 
hIRANOA, Rocky V. 1978. On the reconstructability of syntactic change. Proceedings of the 12th 

International Congress of Linguists, ed by W. U. Dressier et al., 504-7 Innsbruck: 

Innsbrucker Beilrage zur Sprachwissenschaft. 
lilTHUN, Marianne, and Lyle CAMPBELL. 1982 On comparative syntax. Papers from the Third 

International Conference on Historical Linguistics, ed by J. P. Maher et al , 273-91. 

(Amsterdam Studies in the Theory and History of Linguistic Science, Series 4, v 13 ) 

Amsterdam: Benjamins 
Ohala, Manjarl. 1974. The abstractness controversy: experimental evidence from Hindi. Language 

50 225-35 
PEARCE, Elizabeth 1984 On weak pronoun changes In Romance University of flelbourne Working 

Papers in Linguistics 10 101-17. 
RICHTER, Elise 1903. Zur Entwicklung der romanischen Wortstellung aus der lateinischen Halle 

ROHLFS, Oerhard 1954. Historlsche Grammatik der italienischen Sprache und ihrer Mundarten, 3: 

Syntax undWortblldung Bern: Francke 
STEELE, Susan M 1977. Review of J Haiman ( 1974): Targets and syntactic change Language 

53 209-12. 
VINCENT, Nigel 1980 Iconic and symbolic aspects of syntax prospects for reconstruction 

Linguistic reconstruction and Indo-European syntax: Proceedings of the Colloquium of the 

IndogermanischeOesellschaft' 1979,edbyP Ramat et al , 47-68 (Amsterdam Studies in 

the Theory and History of Linguistic Science, Series 4, v 19) Amsterdam Benjamins 


WANNER, Dieter. 1985. Proto-histoire du placement des clitiques en roman Actes du 1 7eme 
Congres International de Linguistique et Ptiilologie Romanes (Aix-en-Provence, 1983), 
2 391-405 Marseilles: Laffltte 

Watkins, Calvert. 1976 Towards PIE syntax. Problems and pseudo-problems. Papers from the 
Parasesslon on Dtachronic Syntax, 305-36. Chicago: Linguistic Society 

WINTER, Werner, 1984. Reconstructional comparBtive linguistics and the reconstruction of the 
syntax of undocumented stages in the development of languages and language families Historical 
syntax, ed by Jacek Ftslak. (Trends In Linguistics, Studies and Monographs, 23.) Berlin et 
alibi: Mouton. 

St^udies in the Lingui stic Sciences 
Volume 15, Number 1, Spring 1985 


Omar Ka 

University of Illinois 

The aim of this paper is to analyze the 
phonological processes taking place in Wolof when 
derivational suffixes are added to verb stems. After 
a presentation of the syllable structure and the 
syllabification principles of the language, the 
syllable-changing rules affecting the segmental 
string at the junction of stems and suffixes are 
discussed: namely, gemination, degemination, glide 
insertion, vowel coalescence and vowel insertion. It 
will be proposed that these rules apply whenever the 
syllabification principles fail to syllabify 
exhaustively the segmental string. In turn, 
syllabification will take place after a phonological 
rule has applied. This suggests that syllable 
structure is assigned at different levels of 
phonological derivations. 

0. Introduction. 

This paper applies an autosegmental theory of the syllable 
to Wolof .1 More precisely, I will try to provide, using various 
proposals regarding the syllable (Clements and Keyser 1981, 
1983; Steriade 1982; Noske 1984; Selkirk 1984), an analysis of 
the phonological patterns occurring in Wolof at the junction of 
a verb stem and derivational suffixes. 

After a brief presentation in section 1 of the underlying 
represenation of Wolof syllable structure and the 
syllabification principles of the language, section 2 describes 
the phonological rules affecting the structure of the segmental 
string when certain suffixes are added to the verb stem. It 
will be shown that the effect of these rules is to change an 
otherwise ill-formed syllable structure in accordance with the 
syllabification priciples of the language. 


1 . W olof Syllable Structure . 

Before analyzing Wolof syllable structure and the syllabi- 
fication priciples of the language, a short outline of the 
segmental, system is necessary. 

The vocalic system is depicted in the following chart: 

Front Central Back 

High i u 

Mid-close e e 6 

Mid-open e a o 

Low a 

In addition to these oppositions, there is a contrast between 
long and short vowels; only e and a have no corresponding long 
vowel . 

The consonantal system distinguishes between simple, 
geminate (also called long or strong) and prenasalized 
consonants. Theses three series of consonants have different 
distributions depending on their position within the stem. In 
stem-initial position, the following simple consonants occur: 

p t c k 

b d j g 

m n n fi 

f s X 

y w 

Prenasalized consonants, geminates and consonant clusters do 
not appear in this position. 2 In medial and final positions in 
the stem, all of the above consonants (plus the prenasalized 
ones) can appear, except p, c, k, d.3 In addition, geminates 
also appear; they ordinarily alternate morphologically with 
simple consonants, but the pairing of geminate and simple 
consonant is not always the expected one. We find the following 
alternations : 

expected alternations 

(1) b ~ bb 

?ub"to close" / ?ubbi "to open" 
yab "to load" / yebbi "to unload" 

(2) t ~ tt 

bet "eye" / bett "to pierce" 

boot "to carry on the back" / botti "to take off the 



(3) .i - ,ij 

tej to close" / tijji "to open" 
muj "end" / mu.i.i "to be last" 

f4) g ~ gg 

teg "to put on" / teggi "to put off" 
sag "honor" / saggi "to uncover' 

( 5 ) w ~ WW 

xew "to be in fashion" / xewwi "to be out of fashion' 
yeew "to tie " / yewwi "to untie" 

(6) y ~ yy 

tayal "to be lazy" / tayyi "to be tired" 

(7) 1 ~ 11 

lal "to lay a sheet" / lalli "to put off a sheet" 
sali "dried and salted fish" / salli '"to lose taste" 

( 8 ) m ~ mm 

lem "to fold" / lemmi '"to unfold" 
xam "to know" / xammee "'to recognize"" 

(9) n ~ nn 

beneen "another" / benn "one" 

sen "garbage " / sanni "to throw away" 

(10) n ~ fin 

saan "to plug" / sanni "to unplug"' 

(11) ft ~ hft 

waft "lower back" / (ne)waM "to show the lower back' 

unexpected alternations 

(12) f ~ PP 

sof "to join" / soppi "to disjoin, to change" 

tofo "younger sibling" / topp "to follow" 

sef "to load" / sippi 'to unload" 

(13) s ~ cc 

SOS "soak" / socc "to rinse " 

fas "to tie" / fecci "to untie"' 

tas "to smash" / (ne)tacc "to be smashed" 


fl4) r ~ dd 

teer "to arrive" / teddi "to depart" 

tur "name" / tudd "to be named" 

jur "to give birth" / juddu 'to be born' 

(15) ~ kk 

dee "to die" / dekki "to resuscitate" 

de "doorstep" / dekk "to live'" 

je "forehead" / jekk "to be first" 

(16) X - qq< 

noox "to push in" / fiuqqi "to pull out" 

In each of the above unexpected alternations, the geminate 
is a stop whereas the corresponding simple consonant is a 
continuant, kk, however, does not have a corresponding simple 
variant in the surface: it alternates with 0. The simple 
consonants f , s , x and the corresponding geminates ss > cc , 
qq are all voiceless; however, it is difficult to make the 
members of the pairs correspond in their point of 
articulation: f is a labio-dental and pp a bilabial, s is an 
alveolar and cc a palatal, x is a velar and qa an uvular. In 
contrast, r and dd are both dental and voiced. 

At this point, a natural question arises as to the 
underlying representation for these alternating pairs of simple 
and geminate consonants (in medial and final positions): while 
the underlying representation is straightforward in the 
expected alternations (for example, in the case of ?ub ~ ?ubbi . 
it is clear that a voiced labial stop is underlying) it is not 
so straightforward in the case of the unexpected 
alternations. Two solutions are possible: (1) consider the 
continuants as underlying; (2) consider the stops as 
underlying. If the first solution is adopted, we will need, in 
addition to a rule (or rules) accounting for the alternation 
between a simple and a geminate consonant, a rule of 
"hardening" that converts a continuant to the corresponding 
stop when the continuant appears in a geminate form: 

(18) /f/ ~ /ff/— * pp 

/s/ ~ /ss/ — •> cc 

/r/ ~ /rr/— > dd 

/x/ ~ /xx/ ^> qq 

However, we will have to solve the problem of the source for 
kk . since the corresponding element is 0; i.e. what continuant 
can underline kk? Consider the following alternations where we 
have underlyingly the same morpheme in each pair: 


(17) dee "to die" / dekki "to resuscitate" 

we 'to connect" / wekki "to separate from" 
je "forehead" / jekk "to be first" 

It would be difficult to derive kk in any plausible way 
directly from 0. In the other cases, a stop would appear when 
a consonant appears in a geminate form. But how can we speak of 
being "geminated" so as to produce, via hardening, kk ? 
An alternative would be to posit a hypothetical consonant as 
underlying : the most likely candidate would be the velar 
continuant x. Unfortunately, that consonant has already an 
alternant as . The continuant underlying the ~ kk alternation 
would thus have to be an abstract one--not one of the actually 
occurring continuants in the language. 

This first solution to the unexpected alternations would 
also require that the hardening of a geminate continuant to a 
stop be accompanied by a shift in point of articulation in some 
cases (e.g. when ss hardens it becomes the palatal cc). This 
complication concerning the point of articulation does not 
count as evidence against the hardening solution since the 
alternative solution discussed below will also require a shift 
in point of articulation in the same cases. 

The second solution would take the underlying 
representation of the unexpected alternations to be a 
stop. Thus the basic alternation would be one involving a 
simple versus a geminate stop, as shown below: 

/P ~ PP/ /c "" cc/ /d ~ dd/ 

/q ~ qq/ /k ~ kk/ 

The geminate stops would undergo no additional rules. The 
simple stops p, d, c and k would have to undergo a rule of 
spirantization which converts them into a continuant (or zero 
in the case of k) . The environment for this spirantization 
process would appear to be post-vocalic (since intial p, d, c 
and k do not undergo the rule). We will assume this environment 
for spirantization in the present paper, but there are a number 
of additional considerations that could conceivably lead in a 
somewhat different direction. It will be this rule of 
spirantization that will be responsible for the fact that the 
simple consonants p, d, c and k do not appear phonetically in 
medial or final position in the stem. Notice also that when a 
simple stop undergoes spirantization, there will also have to 
be a shift in the point of articulation (e.g. when c 
spirantizes, it becomes s). 

We will adopt this second solution on the grounds of 
economy and simplicity: there is no need to posit any 


underlying abstract segment to account for the ~ kk 
alternation. There are some interesting aspects to the 
formulation of spirantization that we will not dwell on 
here. First, notice that we run into problems if we allow k to 
become a continuant first (along with p, d and c) and only then 
delete. Why? Because if k becomes a continuant, it would become 
the velar continuant x and and the x that actually occurs in 
Wolof does not delete in post-vocalic position. Of course, if 
we said that the actually occurring x is really a uvular 
underlyingly (which is why it alternates with ^3 under 
gemination), then we could let k become a continuant, delete 
the resulting velar continuant in post-vocalic position, and 
only then change the underlying uvular continuant to a surface 
velar continuant. But barring this somewhat abstract approach 
to the problem, we will have to delete k directly, by a rule 
such as (19), and then spirantize p, d, and c by a rule such as 
(20). The second issue is how we are to explain why p and c 
(but not t) spirantize and why d (but not b, j, and g) 
spirantizes. It seems as though the rule should be restricted 
to voiceless stops and af f ricates--but why is t not affected 
(it is a voiceless stop) and why is d instead affected? We have 
no answer to this question at the present time and simply 
assume that (20) will have to refer to an unnatural class of 
sounds . 




[-cont] — > [+cont] 

The following derivations illustrate the spirantization 
processes (19) and (20). 

(21) /It sop 1*/ "to join" 

sof spirantization (19) 

(22) /« fac 0/ "to tie" 

fas spirantization (19) 

(23) /« teed »/ "to arrive" 

teer spirantization (19) 


(24) /» deek »/ "to die" 

dee k - drop ( 20) 

(25) /« sopp »/ "to like" 

sopp (19) and (20) are inapplicable 

(26) /tt fecc «/ "to dance" 

fecc (19) and (20) are inapplicable 

(27) /Jf tedd }»/ "to live honorably" 

tedd (19) and (20) are inapplicable 

(28) /« lekk «/ "to eat" 

lekk (19) and (20) are inapplicable 

After this brief presentation of the segmental system, 
let us turn now to the syllable structure of Wolof verb 
stems. On the surface, the following syllable types are met: 

a. CV (or CVCV) ji "to plant" 

fo "to play" 

nuyu "to greet" 

pare "to be ready ' 

b. CVV 

c. CVC 


















put on" 






load ' 









lay (a sheet) 

d. CVCC takk "to fasten" 

gemm "to close eyes' 

depp "to invent" 

samp "to plant" 


CVVC teer "to stop (for a vehicle) 

suul "to bury" 

xaar "to wait for" 

boot "to carry on the back" 

CVCVC raxas "to clean" 

xaraf "to be invited" 
selem "to wash one's face' 

Following Clements and Keyser's (1981,1983) approach to 
the syllable, we will distinguish three tiers in syllable 
representation: the segmental tier, the syllable tier and the 
CV-tier, which is intermediate between the first two. The 
elements of the CV-tier divide themselves into syllable peaks 
and syllable margins. A syllable peak represents any segment 
dominated by a V, and a syllable margin any segment dominated 
only by a C . Thus in lem , [e] represents the syllable peak and 
[1, m] the syllable margins (respectively left and right 
margin) : 

(29) a 

C V C 

In CV terms, the syllable structure of Wolof verb stems obeys 
the following syllabification principles: (1) the syllable 
peak may consist of a short vowel or a long vowel: the contrast 
between them is represented in terms of non-braching 
vs. branching nodes on the syllable tree; (2) the syllable 
right margin is an optional constituent: it may consist of zero 
consonant, one consonant, a geminate or a prenasalized 
consonant, but it cannot be composed of more than two 
consonants (if we consider a geminate or a prenasal as 
representing two consonants in the skeletal slots); (3) each 
syllable begins with a consonant, hence the syllable left 
margin is an obligatory constituent: it consists of only one 
consonant^ ; (4) the syllable types described above show that 
either the syllable peak or the syllable right margin (but not 
both) may branch within the syllable, i.e. syllables of the 
type CVVCC are not met in Wolof on the surface. 

Taking into consideration these syllabification priciples, 
it is possible to formulate an abstract syllable template 
accounting for the possible syllable structures of all verb 


stems at the surface level 


' (gi: 


This syllable template can be viewed as a well-formedness 
condition on the syllable structure of phonological 
represenations in Wolof. 

The stems can be combined with various derivational 
suffixes, 6 such as: 



"directional ' 

-o "reciprocal' 
V V 


-al "causative' 

-al "benef active' 

-at 'iterating" 
V V C 


-le "participant' 

-kat "agent " 

The question of the underlying representation of some of these 
suffixes will be dealt with in section 2. In general, the 
combination of these suffixes with a stem is subject to the 
same syllabification principles as operate for the stem in 
isolation. There is one important difference, however. A stem 
syllable may end in a geminate or a prenasalized consonant when 
that stem syllable is word-final. But a stem syllable may not 
end in a geminate or prenasal when that syllable is not word 
final . 


2 . Phonologic al Rules and S yll able Structure . 

When the suffixes listed in section 1 are combined with 
verb stems, some phonological rules must apply, namely: 

(1) gemination 

(2) degemination 

(3) glide insertion 

(4) vowel coalescence 

(5) vowel insertion 

In this section, we will show that these rules apply 
whenever the general syllabification principles stated in 
section 1 fail to syllabify exhaustively the segmental 
string. We will also give evidence that syllabification 
reapplies after a phonological rule has taken place. This 
implies that syllable structure is assigned at different levels 
of phonological derivations before or after the application of 
each phonological rule. 

2 . 1 The gemination rule . 

Consider the following alternations: 

(32) ?ub "to close" / ?ubbi "to open" 

yab "to load (a vehicle)" / yebbi "to unload (a 

vehicle) '"^ 
tej "to close (a door)" / tijji "to open (a door)" 
teg "to put on" / teggi "to put off" 
lem "to fold" / lemmi "to unfold" 
sof "to join" / soppi "to disjoin, to change" 
sef "to load" / sippi "to unload" 
saf "to be tasteful" / sappi "to loose taste" 
fas "to tie" / fecci "to untie" 
fal "to elect" / folli "to dismiss" 

lal "to lay (a sheet)" / lalli "to put off (a sheet)" 
xew "to be in fashion" / xewwi "to be out of fashion" 

To explain these alternations, a rule of gemination should 
be posited, which - in linear terms - lengthens (or 
strengthens) a simple consonant in stem final position before 
the suffix -i "reversive". It can be formulated as follows in 
standard CV-phonology : 

(33) F F 

c c — > c c 

— ' stem 

so p 

+ i 

1 1 /\ 


C V C C 



Thus, words like lemmi or soppi will have the following 
derivations (via (33)): 

(34) /I e m + i/ 1 e m +i 

1 I I I > I I /\ 1 


(35) /sop +i/ 

: I I I 


Syllabification then applies to the output of the 
gemination rule, making the second C element of the geminate 
the obligatory left margin to the second syllable and making 
the first C element of the geminate the right margin of the 
first syllable. Cf . (36) and (37): 

( 36 ) 1 e m i 
i I /\ I 
C V (i; C V 

( 37 ) so p i 
I 1 /\ 1 

Notice that prior to gemination, the syllabification principles 
would make the stem-final simple consonant the left margin to 
the second syllable and there would be no right margin to the 
first syllable. This example demonstrates that syllabification 
must follow the gemination rule in (33), since the newly 
created C position must be syllabified. We will show below that 
syllabification also applies to the underlying representations 
as well as to the output of phonological rules. 

An alternative description of gemination in Wolof would be 
to analyze -i "reversive" as containing an unassociated C in 
underlying representation: 





When this suffix combines with a stem ending in a single 
consonant, the unassociated C undergoes a rule affiliating it 
to that consonant, from which it takes its features: 


r t e g ] 
C V C 

C V 


( teggi , from 

teg + Ci ) 

Such a process can be seen as a consequence of a general 
principle under which unassociated C-elements or V-elements 
trigger the automatic spreading of adjacent consonants or 
vowels on the segmental string (Clements and Keyser 1983). 

Before discussing the unassociated C approach, let us 
first examine the problem of the treatment of geminate 
consonants in Wolof, since that treatment will play a 
signifigant role in deciding between the two approaches to the 
reversive suffix. Within the non-linear framework, phonological 
structure is conceived of as involving seperate tiers of 
represenation that are linked to each other by association 
conventions: the CV tier which provides the syllable 
information, and the segmental tier which provides the feature 
matrixes. The units of the CV tier are called C or V slots, 
whereas the units of the segmental tier are referred to as 
segments : 


[ F ] 

b. [ G ] 

segmental tier 
CV tier 

A sequence of identical segments is automatically represented 
on the segemntal tier as a single element linked to two 
consecutive slots on the CV tier, following the Obligatory 
Contour Principle (cf. McCarthy 1979, Leben 1980, Kenstowicz 
1982, Steriade 1982): 


1 41) X + Y Z 

i I /\ 

C + C C C (cond. : x=y=z) 

A long consonant is defined as a segment mapped to two C 
slots, and a short consonant is defined as a segment linked to 
just a single slot. Thus, the Wolof stems lem "to fold", saf 
"to join", takk "to tie" will have the following 
representations : 

(42) /I em/ /so p/ /t a k/ 

1 I i ill I I /\ 

cvc cvc cvcc 

(r a 0" 

The adjunction of -i "reversive" to these stems will give, 
after syllabification: 

(43) /lem+i/ lem+i 

Mi I ^ I I /\ I 


/sop+i/ sop+i 

I I I I > I I /\ I 


/tak+i/ tek+i 

II A / ^ I I /\ I 

CVCCV cvcc V 

In accordance with syllabification principle 3 (SPs ) , the 
second C of the geminate becomes the left margin of the 
subsequent syllable. 

The representaions above may be contrasted with those 
obtained after adjunction of the suffix -i "directional" to the 
same verb stems : 


(44) lem "to fold" / lemi "to go and fold" 

to go and load" 















sof i 











"to go and join" 
"to go and elect' 
"to go and tie" 

(45) lem 

C V C V 

t a k + i 
i I A I 
C V C C V 


(In the case of /sop + i/, the rule of spirantization converts 
the short stop to the corresponding continuant since that 
simple stop is preceded by a vowel). 

With this background, let us now return to the analysis in 
which the reversive and directional suffixes are treated 
respectively as i vs . i this analysis 
1 I 


will contrast the suffix "reversive" (which has an unassociated 
C preceding the vowel i) with the suffix "directional" (which 
will not contain an un associated underlying C slot) in terms 
of an underlying phonological difference. This difference in 
underlying structure will produce gemination in the case of the 
reversive suffix but not in the case of the directional 
suffix. For example, (39) above illustrates why (in this 
approach) the reversive produces gemination of a stem-final 
simple consonant. (46) illustrates why the directional does not 
induce gemination in a similar situation. 

(46) [teg] i 


C V C V 


( teei . from: teg + i ) 



The unassociated C analaysis for the reversive suffix 
works very straightforwardly for the cases where the stem ends 
in a simple consonant. The situation becomes a little more 
complex when the stem ends in a geminate or a prenasalized 
consonant underlying. In this case the addition of the 
reversive suffix and the directional suffix produces exactly 
the same output with respect to the stem-final 
consonant--namely , the vowel -i is appended directly to the 
stem without any change in the consonant. Thus, from takk "to 
tie" we get takki "to go and tie" and tekki "to untie" (notice 
that the reversive suffix does induce a change in the vowel of 
the stem in certain cases whereas the directional does not; we 
do not treat problems of vowel quality alternations in Wolof in 
this paper) . 

Given the analysis of the directional suffix as consisting 
just of a vowel on the CV-tier, there is no problem in 
analyzing the directional form of a stem ending in a geminate 
or a prenasalized consonant: the second C of the geminate or 
prenasal will be syllabified with the following vowel. 

(47) [gem] i 

) I /\ I 
C V C C V 

However, if we regard the reversive suffix as beginning with an 
unassociated C slot, some problems do arise when this suffix is 
added to stems ending in a geminate or a prenasalized 
consonant: we will obtain 3 successive C slots on the CV-tier, 
but in the correct phonetic output there must just be two C 
slots. In other words, the unassociated C position in the 
reversive suffix must end up unrealized in the phonetic 
surface . 

We could perhaps achieve this effect by preventing a 
geminate or a prenasal consonant from spreading onto an 
unassociated C--this would make sense in that the language does 
not have triply long consonants nor does it have prenasalized 
geminates. If we thus block spreading of such consonants onto 
the empty C slot, that C will remain unassociated and thus will 
receive no phonetic realization. 

There is, however, another pertinent fact that complicates 
the situation further. Namely, when a stem that ends in a 
geminate or a prenasalized consonant precedes a suffix that 
begins (overtly) with a consonant, there is a rule of vowel 


insertion ( cf . section 2.5 below) that must apply to convert 
the CCC string to CCVC . Thus given an underlying structure 
/takk+Ci/, even if we succeed in keeping the unassociated C of 
the suffix from associating to the features of the stem-final 
consonant, we still must explain why an epenthetic vowel does 
not appear between the stem-final CC and the C-initial suffix, 
as in (48) below: 

(48) *[tak] i *[samp] i 

( I /\ \ I'M ' 


MX \; V \l/ \/ V 

In order to prevent epenthesis here, we would have to delete 
the unassociated C position prior to vowel insertion. But this 
means that in effect the unassociated C slot receives no 
independent conf irmation--i . e. although it accounts for the 
gemination of stem-final simple consonants, it does not behave 
like a consonant in other respects (namely, in inducing vowel 
epenthesis after stems ending in a geminate or a prenasal 
consonant) . 

Another possible approach to deriving tekki from /takk+Ci/ 
exists. Suppose that we did allow the features of the 
stem-final consonant to spread onto the unassociated C slot of 
the reversive suffix. This would mean that the suffixal C 
position would now be part of a geminate construction. We could 
then try to explain the absence of epenthesis in terms of the 
long-observed fact that geminates resist being split up by any 
insertion processes (cf. Hayes 1984, Steriade and Schein 
1984). Given a structure like (49), 

(49) [F] 


it is possible to insert a V slot and obtain 


(50) [F] 


However, that inserted V slot cannot get any features; in order 

for it to be associated with some set of features, the crossing 

of association lines would be necessary, as (51) and (52) 

(51) * [F] fG] * [G] [F] 

ccvc ccvc 

Since autosegmental phonology assumes that such a crossing of 
association lines is universally prohibited, the inserted V 
slot would be predicted to remain unassociated and the lack of 
an overt epenthetic vowel would thus be explained in a 
principled way. Of course, it is not clear exactly what 
phonetic implications are made by permitting a V to remain in 
the CV-tier unassociated with any features. The representaion 

(52) [f] 


still is not a satisfactory one since it predicts, in effect, a 
triply long consonant. In order to generate the correct surface 
form, some rule will be required to delink one of the 
consonants. Since there is no independantly motivated process 
in the language that would take CCC to CC , this represents a 
complication of the grammar that results entirely from the 
assumption that the reversive has an unassociated C slot. 

We have shown that the behavior of stem-final geminates 
and prenasal consonants do not in any way add support to the 
analysis of the reversive as having an unassociated C slot. We 
will show in 2.2 below that there are two other suffixes whose 
behavior is pertinent to the issues involved here and that they 
do not in any way provide support for the unassociated C 
analysis. We will assume in this paper that the difference 
in behavior of the reversive and the directional can not be 
explained in terms of an underlying phonological difference. We 
will assume that both suffixes begin with a vowel and that the 
rule of gemination is morphologically delimited. 

2. 2 The degemination rule. 

We will be concerned here with two suffixes: - al 
"causative" and -o "nominalizing ' . Consider the following data: 



raf et 


"to bve clean" / setal "to clean" 
"to be black" / fiuulal "to blacken" 
"to be beautiful" / rafetal "to make 

"to be could" 
"to follow" 
"to be safe" 
"to be in bed' 
"to come into' 
sonn "to be tired" 

/ serai 
/ tofal 
/ musal 
/ teral 
/ dugal 

'to cool" 
'to add" 
to save" 
'to put in bed " 
'to intoduce" 

/ sonal "to tire, to bother' 




'to follow" / tofo "younger sibling' 
to follow his/her destiny" 

/ reefo "destiny, fate" 
'to be tired" / coono "tiredness" 
'to celebrate, to party" 

/ mbaalo "wine" 

A rule of degemination operates here, converting a 
geminate (or strong, or long) consonant in stem-final position 
into a simple consonant before the suffixes -al "causative" and 
-o "nominalizing" (with vowel-final stems, a rule of glide 
insertion or a rule of vowel coalescence will apply, as will be 
seen in 2.3. and 2.4.). 

The degemination rule can be formulated as follows: 


C C 



Subsequently, the spirantization rule will convert a simple 
stop to the corresponding continuant after a vowel . 

Thus, words likle sonal . tofal will be derived as follows: 


/s o 
I 1 
C V 


c c 





s o 

1 1 

n + a 1 
1 1 1 

1 1 
C V 

C V c 

s o 

1 , 

n + a 1 
1 i 1 

1 1 
C V 

1 1 1 
C V c 

degemi nation 


According to the syllabification principles of the language, 
each syllable peak is provided a left margin, since it is an 
obligatory constituent of the syllable; an optional C at the 
right of a peak is syllabified with that peak and becomes the 
right margin of the syllable. 

(57) /to p + o / 
I 1 /\ I 
C V C C V 

top + o degemination 

I I I I 
C V C V 

t o f + o spirantization 

/ I \ I 
C V C V 

f + o syllabification 

\ I 

C V c 

V \ 

Note here that syllabification could also precede 
spirantization: in each case, we will get the right result. 

The representaions involving -al "causative" may be 
contrasted with those for -al "benef active" : when the latter is 
combined with a verb stem, no phonological rule applies: 






























be safe" 





be tired" 



"to pray for" 

to cultivate for' 

"to buy for" 

"to enter for" 

"to follow for" 

"to be safe for" 

"to be tired for" 

An alternative to the preceding analysis involving a 
morphologically-restricted degemination rule would be to posit 
an unassociated C position before the causative and 
nominalizing suffixes but not before the benefactive 
suffix. For example, the causative would be represented as in 
(59) below. 


Such an alternative would be initially motivated by the fact 
that in many languages degemination occurs before a 
C. Unfortunately, it is not the case in Wolof: we should get 
vowel insertion in this particular enviroment (cf. 2.5.), but 
we do not. Furthermore, we would expect all single Cs in 
stem-final position to be realized as geminates before the 
causative and nominalizing suffixes, but this is not the 
case (cf. setal, nuulal) . Therefore, this analysis would not 
account for the data and must be rejected. Thus we see that 
for the suffixes that cause degimination, a morphological 
solution must be accepted and a phonological solution involving 
an unassociated C slot rejected. This does not, of course, mean 
that a similar morphological solution for the gemination rule 
discussed in section 2.1 is necessary, but it does suggest that 
both gemination and degemination may be alike in having been 

2.3. The glide insertion rule. 

The glide insertion rule is illustrated by the following 
examples : 

(60) ji "to plant" / jiwaat "to plant again" 
( -aat "iterative") 

(61) nuyu "to greet" / nuyuwaat "to greet again' 

f62) woo "to call" / woowal "to call for someone" 
( -al "benef active" ) 

(63) soppi "to change" / soppiwaat "to change again" 

(64) ,ii "to plant" / jiyi "to go and plant' 
( -i "directional") 

(65) woo to call" / wooyi "to go and call" 

(66) fo "to play" / foye "to play with" 
( -e "instrumental") 

(67) woo "to call" / wooye "to call from" 
( -e " locative" ) . 

A glide is inserted between (a) a vowel-final stem and a long 
vowel-initial suffix and (b) an open monosyllable stem and a 
vowel-initial suffix. The quality of the glide is determined by 
the following vowel. We will ignore here the issue of the 
precise mechanism by which the inserted glide consonant gets 
assigned its specification on the front/back dimension. The 
rule of glide insertion will be as in (68): 

V V 


The derivations below illustrate the rule: 

(69) / j i + at/ 
I \ /\ I 
C V V V C 



J i 

w a t 

1 1 

' /\ 1 

C V 

C V V c 

I 1 I /\ I 

C V C V V c 

(70 1 / w c + a 1 

j /\ / / 
C V V V c 

: V V c V c 

f o 

[ , 



1 1 
C V 


f o 

1 1 

y e 

1 1 

1 1 
C V 

1 1 

C V 

f o 

1 1 

y e 

1 1 

1 1 
C V 


t I 

C V 


glide insertion 


w o w a 1 glide insertion 

/ /\ I I ) 
C V V C V c 

w o w a 1 syllabification 

/\ 1 i 

glide insertion 


The glide insertion rule can be conceived of as a hiatus 
breaking rule; it provides a left margin for peaks in 


syllable-initial position since the Wolof syllable template 
rules out syllable-initial peaks. 

As we have seen, glide insertion applies in specific 
enviroments , either between a vowel-final stem and a long 
vowel-initial suffix or an open monosyllabic stem and a 
vowel-initial suffix. In all other cases of vowel 
juxtaposition, a vowel coalescence rule should apply, as will 
be seen in the next section. Crucially, this implies that glide 
insertion should be ordered before vowel coalescence in Wolof. 

2.3. The vowe l coalescence rule. 
Consider the following data: 

(72) nuyu "to greet" -t- -e "instrumental" : 
nuyoo "to greet with" 

(73) pare 'to be ready" + -al "benef active' 
pareel "to be ready for someone' 

(74) tijji "to open" + -e "instrumental" 
tijjee "to open with" 

(75) saaga "to insult"' + -al "causative" 
saagaal '"to cause to insult"" 

(76) songoo "to attack each other" + -al "'causative": 
songool "to cause to attack each other". 

A coalescence rule operates here, collapsing the final vowel of 
a polysyllable stem and the initial short vowel of a suffix; it 
can be formulated as follows: 

(77) [ a hi ] ^ _ _ _ [ -hi ] 

|o(bk / ard / VaTr] [i bk / 'i rd / 9 ATrJ 

r -- -- --, + 

V "" ~ - V 

i.e. the resulting representation will be; 


(78) r -hi 1 


o(bk / ard / VATR J 
V V 

The derivations below illustrate the coalescence rule: 

(79) /nuyu + e/ 

ill! i 

C V C V V 

n u y o coalescence 

I I i /\ 

C V C V V 

n u y o syllabification 

I I I /\ 

C V C V V 

\/ vx 

(80) /pare + al/ 

Mil II 

C V C V V c 

pare 1 coalescence 

i \ \ /\ [ 
C V C V V C 

pare 1 syllabification 

I I » /\ 1 
C V C V V c 

The coalescence rule can be motivated by the syllabification 
principle 1 (SPi ) of Wolof : only a long or short vowel may 
occupy the peak position. Sequences of nonidentical vowels are 
ruled out by this rule, which in effect prohibits them from 
being dominated by one syllable node. Given that a syllable 
cannot begin with a vowel, a sequence of two nonidentical 


vowels can neither be in the same syllable nor in different 
syllables. We seen then that vowel coalescence and glide 
insertion play a similar role in resolving a situation where 
a vowel sequence cannot be syllabified. Vowel insertion 
resolves the problem by providing the second vowel with an 
onset so that this vowel can be syllabified. Vowel coalescence 
takes two nonidentical vowels and makes them into a long vowel, 
thus rendering the second vowel capable of being put in the 
same syllable as the first vowel. 

2.5. The vowel inser tio n ru les. 

There exist two types of vowel insertion rules in Wolof: 
(1) schwa insertion; (2) vowel epenthesis. While schwa 
insertion applies between a stem and a suffix, vowel epenthesis 
operates within a monosyllablic stem. 

2.5.1. Schw a Insertion. 

Consider the following forms: 

(81) lakk "to speak a foreign language" / 

lakk[e]kat8 "speaker of a foreign language' 
( -kat "agent" ) 

togg "to cook" / togg[e]kat "cook". 

The rule of schwa insertion applies in Wolof between a stem 
ending with a geminate or a prenasalized consonant and a suffix 
beginning with a consonant. In autcsegmental terms, if after 
syllabification has taken place in the underlying 
representation a consonant is not part of a syllable, schwa 
insertion will operate and place a schwa after the 
unsyllabif ied consonant. Resyilabif ication will then apply to 
yield the appropriate syllable structure. 

This rule of schwa insertion can be formulated as follows: 

(82) -^ V / C 

The derivations below illustrate the application of (82) 


(83) /la k + kat/ 

I I /\ ill 
C V C C C V c 

la k + kat 

I I /\ \ i I 
C V C C C V c 

syllabification 1 

la k e kat 

i ■' /\ i 111 

cvccv cvc 

schwa insertion 

la k e k a t 

1 I /\ I i I I 

syllabification 2 

After syllabification 1. the second member of the geminate 
remains unassociated. The effect of schwa insertion is to make 

possible the construction of a CV syllable out of the inserted 
vowel and the preceding unassociated consonant. 

2.5.2. Vowel Epenthesis. 

We saw earlier that stem syllables may end in a geminate 
or a prenasal when the stem is word-final. There are also some 
stems that underlyingly end in a consonant cluster that is not 
a geminate or a prenasal. When these consonant clusters appear 
in word-final position, the second member of the cluster is 
unable to be syllabified. A vowel is then inserted in front of 
that unsyllabif ied consonant; this vowel has the same quality 
as the preceding vowel. 

The rule has the following form: 


( 84 ) -> V/ V C C«« 

(85) provides examples of the application of this rule: 

(85) /« xarf «/ 

xarfal "to initiate" (-al "causative") 
xaraf "to be initiated" 

(86) /« sarx «/ 

sarxal "to donate to someone" ( -al "benef active" ) 
sarax "to donate" 

(87) /» selm «/ 

selmu "to wash his/her own face" ( -u "reflexive") 
selem "to wash someone's face" 

We obtain the following derivations: 

(88) /« x a r f ♦*/ 

I I I I 
C V c c 

xarf syllabification 1 

I I I I 

C V c c 

xaraf vowel epenthesis 

I I I I I 
C V C V c 

xaraf syllabification 2 

C V C V c 

V \^^ 

(89) /« s e 1 m «/ 

I I I I 
C V C C 

s e 1 m syllabification 1 

1 I I I 

C V c c 

s e 1 e m vowel epenthesis 

I I I i I 
C V C V c 

s e 1 e m syllabification 2 

C V C V C 
V \^ 

The epenthesis rule plays the role of a cluster breaker; it 
allows an unassociated consonant to be syllabified. 

3 . Conclusion 

This paper demonstrates the importance of a multilinear 
view of phonological structure for lexical derivation in 
Wolof . In particular, Wolof confirms the general notion that a 
variety of phonological processes are best explained if they 
are seen as being caused by the failure of certain CV elements 
to be syllabified after the syllabification priciples of the 
language have applied to the underlying representation; 
phonological rules such as vowel coalescence, glide insertion 
and vowel insertion change the CV structure in such a way that 
the previously unsyllabif ied CV elements can now be 
syllabified. The paper provides evidence that syllabification 
does not take place only in the underlying representation in 
Wolof: it can both precede and follow the application of 
phonological rules. While other facets of the multilinear 
approach to phonology may possibly be invoked (for example, 
unassociated C slots, the Obligatory Contour Principle) to 
explain aspects of Wolof phonology, the evidence for these is 
not as yet entirely convincing. 



+ This paper has profited a great deal from extensive 
discussions with Charles W. Kisseberth. We are also indebted to 
Michael J. Kenstowicz and Diana B. Archangel i for their 
comments and suggestions on earlier versions of this paper. 

1 The Wolof language is spoken in West Africa, mainly in 
Senegal and the Gambia (now Senegambia); it belongs to the 
Northern West Atlantic subgroup of the Niger-Congo family, 
along with Pulaar (or Fulani) and Seereer ( cf . Greenberg 
1963) . 

2 Diachronically , the voiceless prenasals are analyzed as 
having lost their initial nasal segment in that position: 











^P. c, k, d appear in these positions only in borrowed or 
derived words. 

4 In the orthography of Wolof, aa is simplified into g. 

5 Cluster-initial words are ruled out by epenthesis or 
prothesis; this is clear particularly in the borrowings 
from French: 

classe kalaas 

statue estati 

gris giri 

sport espoor 

6 There are 40 derivational suffixes, but no derivational 
prefixes in modern Wolof (cf. Ka 1981); we list here only 
those relevant to our concerns here. 

''A subsequent rule changes the nature of the stem vowel in 
the following way: 

'^.. /" 

8 The inserted vowel e is not noted in the orthography of 
the language. 


CLEMENTS, G., and S. KEYSER. 1981. A three-tiered theory of the 

syllable. Occasional Paper 19, The Center for Cognitive 

Science, MIT. 
CLEMENTS, G., and S. Keyser. 1983. CV-Phonology . Linguistic 

Inquiry Monograph. Cambridge, MA:MIT Press. 
DONEUX, J. 1975. Quelle Phonologie pour le Wolof? Dakar: CLAD 

GOLDSMITH, J. 1976. Autosegmental Phonology. Doctoral 

Dissertaion. Cambridge, MA: MIT. 
GREENBERG, J. 1963. The Languages of Africa. The Hague: Mouton. 
HAYES, B. 1984. Inalterability in CV-Phonology. Unpublished 

paper. Los Angeles: UCLA. 
KA, 0. 1981. La derivation et la composition en 

Wolof. Dakar: CLAD 77. 
KENSTOWICZ, M. 1982. Gemination and spirantization in 

Tigrinya. Studies in the Lingustic Sciences, 

12:1. 103-122. 
LEBEN, W. 1980. A metrical analysis of length. Linguistic 

Inquiry 11: 497-509. 
McCarthy, J. 1979. Formal problems in Semitic phonology and 

morphology. Doctoral Dissertation. Cambridge, MA: MIT. 
McCarthy, j. 1984. OCP effects: gemination and 

antigemination. Unpublished MS. University of Texas at 

NOSKE, R. 1984. Syllabification and syllable changing rules in 

French. In H. van der Hulst and N. Smith, eds . : The 

structure of phonological representations, part II, 

pp. 257-310. Dordrecht, Holland: Foris Publications. 
SELKIRK, E. 1984. The syllable. In H. van der Hulst and 

N. Smith, eds. : The structure of phonological 

representations, part II, pp. 337-383. Dordrecht, 

Holland: Foris Publications. 
STERIADE, D. 1982. Greek prosodies and the nature of 

syllabification. Doctoral Dissertation. Cambridge, MA:MIT. 
STERIADE, D. , and B. Schein. 1984. On geminates. Unpublished 

paper. Cambridge, MA: MIT. 

studies in the Linguistic Sciences 
Volume 15, Number 1, Spring 1985 


Yamuna Kachru 

In this paper, several issues have been raised with respect 
to standard paradigms of research in second language acquisition 
from the perspective of multilingual societies of the non-Western 
world. Data from institutionalized varieties of English have 
been brought to bear upon explanations in terms of interlanguage 
and fossi 1 ization. It has been argued that if discourse considerations 
are responsible for the non-nativeness of institutionalized varieties, 
as has been claimed in several recent studies, the non-native 
features can hardly be characterized as 'errors' and explained 
away as due to fossi 1 ization, overgeneral ization, ignorance of 
rule restriction, etc. A number of questions have also been raised 
about the so-called 'approaches' to, or methodologies of, language 
teaching currently in favor on both sides of the Atlantic. It 
has been demonstrated that second language acquisition research 
needs to take into account the research findings of sociol inguistics 
in language and social identity, and bi-/multi 1 ingual ism. Similarly, 
the research in teaching methodologies needs to be sensitive to 
the wider context of language teaching. Unless the data base 
of research in these areas is expanded, the claims to universality 
of research findings in second language acquisition and language 
teaching methodologies will remain suspect for most of the non- 
Western world. 


The field of foreign and/or second language teaching is so vast 
that it may be useful, at the outset, to indicate the exact context 
of my discussion. First, I will not make any distinction between foreign 
and second language teaching: This distinction is not very clear- 
cut from the perspective of a majority of non-Western countries. I 
will return to this point later. Secondly, as the issues arise, I 
will refer to second language acquisition, second language learning, 
and second language teaching, as all three are related in the context 
of language education. Thirdly, I will confine myself to posing some 
questions for applied linguistics and language teaching. I am particularly 
concerned with the theoretical framework in second language acquisition 
research, and the methodology of second language teaching. 

Second vs. Foreign Language 

First, let me address the question of second vs. foreign language. 
It is true that historically speaking, for example, English and French 
are foreign languages in several parts of the world where they were 
introduced by the colonial powers. The countries where these two 
languages are used in the present post-colonial era are referred to 
as Anglophone and Francophone, respectively. In these nations, English 


and French are no longer foreign languages, they are used intranational ly 

for purposes such as administration, education, and legal services. 

Hence, they are the most prominent second languages in these countries. 
Nations where English has become a prominent second language are listed 
in 1 below. 

(1) Non-English mother tongue countries where English has official 

Botswana Nauru 

Burma Nigeria 

Camroon Pakistan 

Ethiopia Philippines 

Fiji Sierra Leone 

Gambia Singapore 

Ghana *South Africa 

India Sri Lanka 

Israel Sudan 

Kenya Swaziland 

Lesotho Tanzania 

Liberia Tonga 

Malawi Uganda 

Malaysia Western Samoa 

Malta Zambia 

Mauritius Zimbabwe 


(Fishman, Cooper and Conrad 1977:10,12] 

(*The language situation Is quite complex in South Africa, 
but is not relevant to our discussion.) 

What Is true of English In the above countries is true of French 
in the Francophone countries of Africa. In the following countries 
of Africa, French is the medium of education and hence, of administration, 

(2) Former French colonies where French is the medium of education: 

Algeria Mali 

Benin Mauritania 

Burundi Morocco 

Central African Republic Niger 

Chad Rwanda 

Congo Senegal 

Djibouti Togo 

Gabon Tunisia 

Guinea United Rep. of Cameroon 

Ivory Coast Upper Volta 

Madagascar Zaire 

(Bokamba 1984) 


What is true of English and French in the countries listed in 
1 and 2 above is true of Spanish and Portuguese in Latin America. 
Since I am most familiar with the Anglophone parts of the world, my 
subsequent discussion will focus on English in non-native contexts. 
I will particularly concentrate on English as a Second Language (ESL) 
as a representative case of second/foreign language teaching. This 
is perfectly justifiable on the grounds that a great deal of research 
in the area of Second Language Acquisition (SLA) is devoted to ESL 
al 1 over the world. 

SLA Research: The State of the Art 

As regards the paradigms of research in the field of second language 
acquisition, this area has been approached from four major standpoints 
in the past three decades. The first approach took the position that 
one's first or native language either helps or hinders one in learning 
a subsequent language. Therefore, a careful comparison of the structures 
of the native and target languages is essential for effective language 
teaching. This approach, known as the CONTRASTIVE ANALYSIS HYPOTHESIS, 
was advocated by such well-known linguists and language educators as 
Charles Fries and Robert Lado (Fries 1945 and Lado 1957). A number 
of contrastive analyses of well-known languages appeared and it was 
taken for granted that materials based on them would lead to better 
success in language learning. Good examples of such contrastive analyses 
are the works published by the University of Chicago Press on Spanish- 
English (Stockwell and Bowen 1965 and Stockwell, Bowen, and Martin 
1965), German-English (Moulton 1962 and Kufner 1962) and Italian-English 
(Agard and Di Pietro 1965, 2 vols.) The bibliographies published by 
the Center for Applied Linguistics list several contrastive studies 
involving almost all major languages of the world (e.g.. Gage 1961, 
Hammer and Rice 1965; see also Dechert, Bruggemeier and Futterer 1984). 

Soon, however, disillusionment set in and experienced language 
teachers as well as researchers began to point out that contrastive 
analysis had limited predictive value. It was argued that simply on 
the basis of a comparison of the native and target languages, teachers 
will not be able to identify what causes most difficulty in learning 
the various sounds, words, and sentence patterns of a given target 
language. The errors that the learners make are not always what contrastive 
analysis predicts: It is not always the case that the errors made 
by the learners have their source in their native languages (e.g., 
Lee 1968, Duskova 1969). Researchers also pointed out that some of 
the errors learners make are similar to, or even identical with, the 
errors made by children learning the target language as their first 
language (e.g., Ravem 1968 and 1974). Emphasis thereafter naturally 
shifted to the learner errors. Studies by Corder (1967 and 1971), 
Dulay and Burt (1974) and others pointed out that systematic errors 
provide clues to the progress that learners make in their learning 
task. Hence, ERROR ANALYSIS is more relevant as compared to contrastive 
analysis as the paradigm of research in second language learning (e.g., 
Dulay and Burt 1974). 

The emphasis on learner-centered approaches soon resulted in a 
more comprehensive framework for studying second language learning 


or acquisition. The new approach incorporated the techniques of con- 
trastive analysis and error analysis and became known as the INTERLANGUAGE 
HYPOTHESIS (Selinker 1972). This hypothesis stipulated that systematic 
learner errors provide clues to the process of learning. A periodic 
study of such errors, and a comparison of learner performance in the 
target language with the native and target language systems, will identify 
the successive stages of learning. At each stage, learners have an 
interl anguage system that is different from their native as well as 
the target language system that they are attempting to acquire. A 
learner progresses through several stages of interlanguage before acquiring 
competence in the target language. In a majority of cases of adult 
learners,nati ve-1 ike competence in the target language is difficult 
to achieve. Even at the most advanced stages of the interlanguage, 
adult learners have traces of fossil ization of their native language, 
or of an interlanguage, features in their target language system. 
The interlanguage hypothesis, incorporating the insights of earlier „ 
approaches and concepts such as fossil ization, is by now well-established. 

The interlanguage hypothesis represents a general acceptance of 
the assumption that second language learning is similar to first language 
acquisition (Dulay and Burt 1976). This in turn has led to investigations 
of learner errors in terms of language universals. It is claimed that 
an explanation for errors in learner performance can be found if it 
could be established that it is the marked structures of the target 
language that cause learning problems (Eckman 1977). An example of 
this phenomenon is the following: Relative clause formation in English 
is marked in the sense that it involves, in addition to the use of 
appropriate relative pronouns, the fronting of the relative pronoun 
and hence a change in word order within the relative clause. For instance, 
in the sentence I would like to see the book which you recently bought , 
the relative pronoun which is understood as referring to the object 
of the verb buy , and yet the word order is not you bought which . As 
a consequence of the front shifting of the relative pronoun, the structure 
becomes complex and difficult to acquire from the point of view of 
a learner. As this hypothesis is attractive to researchers interested 
in linguistic universals, it has been adopted by a number of them. 

Recent Trends in SLA Research 

This emphasis on linguistic structure to explain language learning 
was not acceptable to all. A number of researchers proceeded to demon- 
strate that language learning and teaching do not involve language 
structure only. Rather, successful language learning involves competence 
in communicating one's ideas, beliefs, needs, etc., in various types 
of linguistic interactions. 

The activities of the Council of Europe dealing with the problem 
of teaching European languages to immigrant workers in Western Europe 
and Britain ultimately resulted in the development of a new approach 
labelled COMMUNICATIVE LANGUAGE TEACHING. According to this approach, 
what language teaching and learning have to deal with is the use of 
language in social interactions rather than a mere mastery of the 
skills of pronunciation, grammatical structures, vocabulary, etc. 
The main issue in teaching is how to equip the learners with the capa- 
bility to use the target language appropriately in various social situations 
to achieve their communicative goals. This approach and related methodology 


were first proposed in Wilkins (1976) and later elaborated in Munby 
(1978). By now, the communicative approach to language teaching has 
gained wide acceptance on both sides of the Atlantic. It should, 
however, be noted here that the underlying theoretical and methodological 
insights for these approaches were provided by linguists and socio] inguists 
such as J. R. Firth, Dell Hymes, and M. A. K. Halliday. 

SLA: A Non-Western Perspective 

Approaching second language teaching and learning from a non- 
Western perspective opens up a whole new range of issues not normally 
addressed in the approaches discussed above. The issues of literacy 
and language in education are inextricably tied in with the issue of 
language teaching and learning in those areas of the world where English, 
French, Portuguese, and Spanish are the media of education without 
necessarily being the native languages. In Ivory Coast, Kenya, Singapore, 
and South Asia, to name just a few nations, it is not a question of 
teaching French or English as a second language, but a question of 
teaching literacy skills, mathematics, sciences, history, etc., through 
French or English. Thus, the whole issue of teaching a second language 
is linked with questions of language policy and planning. (Se e ARAL 
4, 1983 for a discussion of literacy in several regions of the world, 
and Rubin and Jernudd 1975 and Cobarrubias and Fishman 1983 for questions 
of language policy and planning in developed and developing countries.) 

It is worth noting that in a majority of the nations of the non- 
Western world (e.g., the ones listed in 1 and 2 above), it is not a 
question of ' bi 1 ingual ism.or not' (Skutnaab-Kangas 1984), as is clear 
from the data in 3 below. 

(3) Language profile of selected countries of the non-Western 

a. India: Number of mother tongues reported in the census: 
1652. These belong to four language families: 
Indo-Aryan, Dravidian, Austro-Asiatic, and 
Sino-Tibetan. Official languages: Hindi 
and English. Media of higher education: 
sixteen major languages and English. 

Kenya: Four major languages: Swahili, Gikuyu, 

Lubya (Bantu family), and Luo (Nilotic family). 
Official languages: Swahili and English. 
Medium of higher education: English 

Singapore: Three major ethnic groups: Chinese, Malay, 

and Tamil. Languages: Several Chinese dialects, 
Malay, and Tamil. Official languages: Mandarin, 
Malay, Tamil, and English. Medium of higher 
education: English. 

It is clear from the above that a majority of the population in 
these countries is bi-/multil ingual and has been for centuries. Thus, 
concerns of bi-/multi 1 ingual ism are extremely relevant for research 
on second language acquisition from the point of view of these countries. 


The Case of Non-Native Varieties of English 
I would like to elaborate on these concerns with one example. 
The case in point is that of English around the world. In many of 
the countries where English is used either as an official language, 
as a language of higher education, or for international trade and commerce, 
diplomacy, etc., varieties of English have developed which are not 
identical with the native varieties used in Australia, Britain, Canada, 
New Zealand and the United States of America (see Bailey and Gorlach 
1982, B. Kachru 1982, 1983, Piatt, Weber and Ho 1984, Smith 1983, among 
others, for details regarding these varieties). In some of these varieties 
there is a considerable body of creative literature. The perceptions 
of some of the users of these varieties is given in 4 below. 

(4)a. Most Singaporeans recognize the fact that they speak English 
differently from the so-called "native speakers" of English. 
... They accept these differences but are quite content 
to speak English their "own" way as long as they can be understood 
by fellow-Singaporeans and foreigners. 

(Richards and Tay 1981 :54) 

b. I feel that the English language will be able to carry the 
weight of my African experience. But it will have to be 

a new English, still in communion with its ancestral home 
but altered to suit its new African surroundings. 

(Achebe 1965:30) 

c. I am an Indian, very brown, born in 
Malabar, I speak three languages, write in 

Two, dream in one. Don't write in English, they said, 

English is not your mother-tongue. Why not leave 

Me alone, critics, friends, visiting cousins. 

Everyone of you? Why not let me speak in 

Any language I like? The language I speak 

Becomes mine, its distortions, its queernesses. 

All mine, mine alone. It is half English, half 

Indian, funny perhaps, but it is honest. 

It is as human as I am human, don't 

You see? It voices my joys, my longings, my 

Hopes, and it is useful to me as cawing 

Is to crows or roaring to lions, it 

Is human speech, the speech of the mind that is 

Here and not there, a mind that sees and hears and 

Is aware. Not the deep, blind speech 

of trees in storm or of monsoon clouds or of rain or the 

Incoherent mutterings of the blazing 

Funeral pyre. ... 

(Das 1980:38-39) 

Some of the linguistic features that make these non-native varieties 

different from the native varieties of English are given in 5 below. 

(See B. Kachru 1982, Piatt, Weber and Ho 1984, Smith 1981, among others, 
for detai Is) . 


(5) Phonology: 

Different stress placement in words (the syllable preceding 
is stressed) : ~~ 

a. Filipino: laborato'ry , chara ' cterized , circu 'mstances 

b. Singaporean: facu ' Ity , educa' ted , conte'xt , prefere'nce 

c. Indian: de' velopment , chara' cter 

d. Nigerian: su 'ccess , recogni ' ze , investiga' te 

(lowenberg 1984b) 

Lex icon: 

a. Singaporean: Handicaps on our island republic get stares 

wherever they go. 

(Lowenberg 1984b) 

b. Indian: What are the subjects you offered at B.A.? 

(Lowenberg 1984b) 

c. Ghanian: He does not use a chewing stick to clean his teeth. 

(Lowenberg 1984b) 

d. East African: He overl istened to the boy's conversation. 

(Hancock and Angogo 1982:318) 

S yntax : 

A. Countability of non-count nouns: 

a. Filipino: He has many luggages . (Gonzales 1983) 

b. Singaporean: Give me a chalk . (Lowenberg 1984b) 

c. Nigerian: I lost all my furnitures and many valuable properties. 

(Bokamba 1982:82) 

d. Indian: There are historical as well as synchronic evidences 

which can support separating of aspiration from stops. 
(IL 35:3, 1976:230) ^ 

B. Resumptive pronouns: 

Arab: the time I spent j_t in practice 

Chinese: We put them in boxes we call them rice boxes. 

(Schachter 1976) 
Nigerian: The politicians and their supporters, they don't 
often listen to advice. 

(Bamgbose 1982:106) 


a. Singaporean: Are you feeling lonely, bored or having no time 

to get friends? 

(SM July 7, 1984:5) 

b. Indian: You are all knowing , friends, what sweetness is in 

Miss Pushpa. 

(Ezekiel 1976) 

Interlanguage or bilingual 's creativity? 

The above examples and similar data from non-native varieties 
of English give rise to several questions. The first question is 


whether the differences observable in the data are due to overgenerali- 
zation of target language features or transfer from the native languages. 
The difficulty is that this question is not easy to answer. To take 
one example, there is no consistent semantic basis for marking the 
count/non-count distinction in English nouns, especially in the case 
of collective and abstract nouns. In such cases, learners simply follow 
the conventions of their own native languages (cf. examples in A above). 
Similarly, in the case of resumptive pronouns (cf. examples in B above), 
Schachter and Celce-Murcia (1980) argue that such structures in the 
performance of Chinese and Japanese learners are motivated by the topic- 
comment structure of their native languages. Hatch (1978b) claims 
the same about the use of articles (a^, aji, and the) in the English 
of Spanish speakers. In Hatch (1978b), it has been pointed out that 
an analysis of total texts produced by Spanish speakers reveals the 
fact that these learners follow the Spanish convention of use of indefinite 
and definite articles in their English. This learner strategy leads 
to fewer errors in the use of the , but a greater number of errors in 
the use of a/an. 

The following examples from various localized forms of English 
provide further support for the claim that learners follow the discourse 
conventions of their native languages which results in their using 
specific grammatical devices of English in a non-native fashion." 

Discourse : 

a. Indian: 

... The position has belonged to such actresses who come to 
personify, at any given moment, the popular ideal of physical 
beauty . . . 

(H' September 30, 1983:39) 

... They are brought up in such an atmosphere where they 
are not encouraged to express themselves upon such subjects 
in front of others . . . 

( HLI : 194-195) 

The use of such as a correlative of who and where in the above 

examples reflects the conventions of use of cohesive ties (Halliday 

and Hasan 1976) in Indian languages such as Bengali, Hindi, Marathi, 
Punjabi , and others. 

This leads to a further question: if the features identified 
as unique to non-native varieties of English are motivated by discourse 
considerations, as has been demonstrated in studies such as Chishimba 
(1983), B. Kachru (1982, 1983, 1984), Y. Kachru (1982, 1983, 1984), 
Lowenberg (1984a), Magura (1984), among others, then how can they 
be considered instances of fossi 1 ization? How can we distinguish cases 
that exemplify discourse strategies from cases that provide evidence 
for fossi 1 ization? What theoretical justification, if any, is there 
for characterizing features of non-native varieties as fossi 1 ization 
and of the varieties themselves as interlanguages? Which characteristics 
of the non-native varieties, as encountered in creative literature 


or mature writing (i.e., by journalists, critics, authors, etc.), are 
to be treated as illustrations of bilingual's creativity as opposed 
to fossi lization, overgeneral ization, or ignorance of rule restrictions? 
These questions are serious; they cannot be swept under the rug. 
As has been stated above, most of the institutionalized non-native 
varieties are being used in their respective regions as media of higher 
education, administration, and for social interaction. To label them 
interlanguages denies vast populations of these countries a legitimate 
language for conducting their business. 

Obviously, the question of a model of English for education and 
other purposes is crucial for the non-Western world (B. Kachru 1976 
and 1982). As far as the users of the non-native varieties themselves 
are concerned, they are not in favor of a 'foreign' model, as is clear 
from the following tables: 

(6) a. Variety of English presently spoken by educated speakers: 

Singaporeans Indians Thais 

^0 % % 

1. British 40.5 27.4 6.5 

2. American 6.0 3.2 28.1 

3. Australian 0.6 0.0 o'o 

4- Unique 42.3 50.6 40.'3 

5- Others 10.6 18.8 25.1 

b. The variety that we should learn to speak: 

Singaporeans Indians Thais 
% % % 

1. British 

2. American 

3. Australian 

4. Own way 

5. Others 

(These results were obtained in a survey conducted among final 
year Bachelor degree students in Singapore, Hyderabad (India), 
and Bangkok (Thailand). There were 170 Singaporean, 342 Indian 
and 313 Thai students.) 

c- Indian graduate students' self-labeling of their English: 
Identity marke r % 

American English 2.58 

British English 29.11 

Indian English 55.64 

'Mixture' of all these 2.99 

I don't know 8.97 

"Good" English .27 

(B. Kachru 1976:232) 


28.5 49.1 


12.0 31.6 


0.3 0.3 


47.4 3.5 


11.8 15.5 

(Shaw 1981:119-120 


(Kachru 1976 presents the results of a survey carried out in 
India that involved 700 Bachelor and Master's degree students 
in English, and 196 members of faculty and 29 heads of departments 
of Engl ish. ) 

It is clear from the above data that unlike the countries where 
English is used only for international purposes (e.g., Thailand), the 
institutionalized variety users prefer to characterize their English 
as their "own" rather than to conform to some "native" English norm. 
The tables in 6 support the sentiments expressed by scholars and creative 
writers in 4 above. 

Communicative needs and the uses of English 

Looked at from the point of view of communicative needs of the 
users of the localized forms of English, it is clear that the adoption 
of these varieties as models for teaching and learning in their respective 
regions is entirely justifiable. The differences that these varieties 
exhibit serve specific sociocultural needs such as satisfying certain 
conventions of linguistic interactions, whether through an oral or 
written mode. The following excerpt from an Iraqi news report demonstrates 
this dramatically. 

(7) In the name of God, the merciful, the compassionate. 

Great Iraqi people, sons of the glorious Arab nations, it has 
been known to us from the beginning that many parties local 
international, were and still are behind the eagerness of 
the backward and suspect Iranian regime to stir up the dispute 
with, and conduct aggression against and begin the war against 

(from B. Kachru 1982:340) 

The above are the opening paragrapns ot an official statement 
about the destruction of the Iraqi Osirak nuclear reactor by the Israeli 
forces in June 1981. The point of the story - the attack by Israeli 
forces - is mentioned in one sentence after five such short paragraphs. 
Such elaborate build-ups before coming to the point of a story is not 
unique to Iraqi Arabic. To quote Chishimba, "In the cultures of Africa, 
loquacity, ambiguity, redundancy, obscurity and other strategies of 
verbal discourse are markers of wisdom, age, knowledgeabil ity, sex, 
and other socially relevant criteria." (Chishimba 1982:246-247). 

What is suggested is that the unique features of non-native varieties 
deserve to be treated as evidence for bilingual 's creativity rather 
than as evidence for fossil ization (a la Selinker 1972), ignorance 
of rule restrictions, deficiency, etc. In cases where such features 
occur in literary texts, we have less difficulty in accepting them 
as stylistic innovations (Nelson 1984a, 1984b). In the case of expository 
prose or ordinary speech, however, there is an attitudinal factor that 
labels such innovations "un-English." Considering the range of variation 
in dialects within a native English-speaking country, and in varieties 
across different native English-speaking countries, it is not unreasonable 
to suggest that certain features of non-native varieties be accepted 
as legitimate variations. After all, the non-native variations in 
8 below are no more severe than the native ones. 


(8) British: Have you had your holiday yet? 
American: Did you have your vacation yet? 

(Strevens 1977:149) 

British: different from, to 
American: different than 

(Strevens 1977:150) 

Singaporean: So you have to go turn by turn. 

(Piatt, Weber and Ho 1983:48) 

African: ...we are seven and a half million strong and quite 
a number of these can not get jobs to do, so we 
should cut down on bringing forth. 
(Bokamba 1982:88) 

Indian: The concept of idiolect I do not know if people 
still talk about it. 

(rr 35:3, 1974:229) 

This entails a new theoretical framework for research that starts 
with the assumption that people learn languages in order to fulfill 
certain communicative needs which may not coincide with the needs of 
the native speakers of the target language. Consequently, second language 
users develop their own strategies which result in differences at 
each level of the target language structure as well as conventions 
of its use. Second language acquisition research thus has to take 
into account the findings of research in bi-/multil ingualism. 

Issues in methodology 

As regards the question of methodology, several methods have been 
proposed, adopted for a short time, and discarded as the fashions change 
in second language pedagogy (Richards 1984). Very little empirical 
evidence is available to support the claims of effectiveness for any 
particular method, and yet, scarce resources continue to be invested 
in following 'the trend' in ESL classrooms. In the eighties, there 
is a definite shift from the audio-lingual method to the communicative 
approach in the classroom, but, unfortunately, neither approach, as 
currently conceptualized, takes any notice of the situation in non- 
Western countries. The following is typical of many of the countries 
listed in 1 and 2 above. 

(9) Teaching English in Gambia: 

a. General teacher information: 

Gender Qualifications Qualifications by gender 
67% male 27% qualified 28% men qualified 
33% female 73% unqualified 25% women qualified 

b. Bi-/multil ingual ism: 

Average teacher speaks 2.8 languages, one of which is English. 

34% bilingual, 48% trilingual, 15% speak four languages, 
3% speak five languages. 


c. Patterns of language use: English used for banking; 

in linguistic interaction with the head teacher and other 
teachers; in teaching mathematics, sciences, social 
studies; in praising children for their performance; 
occasionally in interacting with the parents of children; 
occasionally in interacting with one's spouse, children 
and friends. 

(Bowcock 1984) 

There is an urgent need for research in the area of suitable method- 
ology for language teaching in crowded, sparsely equipped classrooms 
as compared to the type of classrooms we are familiar with. Recently, 
a group of English teachers and teacher trainers from selected non- 
Western countries visited a number of TESL programs at U.S. universities. 
Their typical concerns were as follows: 

(10) Pakistan: (college-level teaching) 

Two of her biggest problems are large classes (100-200 students) 
and lack of sophisticated resources. She would like to learn 
as much as she can about strategies for teaching large classes 
and where to find (or how to make) inexpensive visual aids. 

Sudan: (high school teacher training) 

60-80 students are often in one class; what can be learned 
on this trip to help teach in this environment? 

One can always take the position that these are impossible situations 
and ignore the whole question. As applied linguists, teacher trainers, 
teachers, and educators, however, I hope we accept the challenge instead. 


In conclusion, we need serious, basic research that will lead 
us to adequate descriptions of English and other languages of wider 
communication around the world in their varied sociol inguistic contexts. 
This has to be accompanied by applied research in teaching methodology 
and curriculum and materials development. At the present state of 
our knowledge regarding what makes second language learning possible, 
it is more useful to encourage different methodologies, both tried 
and familiar methods as well as new ones, rather than to throw out 
any as being out-dated. As suggested in Oilier (1981), different methods 
and classroom practices utilize different areas and different pathways 
of the brain and result in better success in learning. From a non- 
Western perspective, these are the challenges that applied linguistics 
and foreign language pedagogy face today. 

Theoretically speaking, a distinction is made between second 
language learning and second language acquisition. Second language 
learning is said to be a conscious process that involves instruction 
whereas second language acquisition is characterized as a natural, 
unconscious process. Learning and acquisition are both learner-centered 


as opposed to teaching, which is teacher-centered and does not take 
into account factors related to learners such as age, attitude, motiva- 
tion, the difference between input (provided in the classroom) vs. 
intake (internalized by the learner), etc. There is, however, some 
doubt as to whether the distinction between learning and acquisition 
is so clear-cut (Diller 1981). 

See Sridhar (1980) for an insightful discussion of contrastive 
analysis, error analysis, and interl anguage. 

This IS clear from recent publications meant for language teachers 

(and teacher trainers), e.g., Widdowson (1978), Brumfit and Johnson 

(1979), Finnochiaro and Brumfit (1983), and Savignon (1983). 

'Bi 1 ingual ism or not' is the main title of Skutnaab-Kangas (1984) 
which contains a detailed discussion of the problem of minority education 
in Europe. The need for guest workers or immigrant laborers in the 
industrialized nations of Europe has created a situation where it is 
becoming increasingly obvious that the immigrant workers and, more 
importantly, their children have to have access to bilingual education 
if these nations are to avoid a great deal of social and political 

In addition to the sources listed in the References the following 
have provided additional data discussed in this study: 

HLI = Singh, Amrik and P. G. Altbach, eds. 1974. The higher 
learning in India . Bombay: Vikas Publishing House; _I_L = Indian Linguis - 
tics , the journal of the Linguistic Society of India; and n_ = India 
Today , a bimonthly magazine, comparable to Time . The quotes in this 
paper are from the overseas edition. 

I do not mean to suggest that all attested differences between 
native and non-native varieties are motivated by discourse considera- 
tions. Obviously, non-native varieties, too, just like the native 
varieties, have a range of dialect variation (e.g., basilect, mesolect, 
and acrolect in Singapore discussed in Lowenberg (1984a)). Also, in 
any body of attested data, it is likely that there will be a number 
of 'mistakes', whether the data is from a native or a non-native variety. 
What I am concerned with here is the variation that is due to discourse 
considerations. Most such innovations in non-native varieties result 
from restricting or extending the domains of specific devices of English, 
e.g., in the example in 5 under discourse , Indian English extends the 
function of such to a correlative of the relative pronouns who and 
where . It is worth remembering that such does function as a correlative 
in the constructions such as and such that in native varieties, too. 

See Davidson (1980) for a description and illustration of various 
methods practiced currently in the ESL classrooms. 


ACHEBE, Chinua. 1965. English and the African writer. Transition 


AGARD, F. B. and R. DiPietro. 1965. The sounds of English and Spanish. 

Chicago: University of Chicago Press. 
AGARD, F. B. and R. DiPietro. 1965. The grammatical structures of 

English and Spanish. Chicago: University of Chicago Press. 
BAILEY, R. and M. Gorlach. 1982. English as a world language. Ann 

Arbor: University of Michigan Press. 
BAM6B0SE, Ayo. 1982. Standard Nigerian English: issues of identifi- 
cation'. In B. Kachru, ed., 1982: pp. 99-111. 
BOKAMBA, Eyamba. 1982. The Africanization of English. In B. Kachru, 

ed., 1982; pp. 77-98. 
BOKAMBA, Eyamba. 1984. French colonial language policy in Africa 

and its legacies. Studies in the Linguistic Sciences 14:2.1- 

BOWCOCK, Diane. 1984. Teacher attitudes about language usage in 

Gambian schools. Paper presented at the Conference on Language, 

Literacy and Communications in Africa, Fifth Annual Roundtable 

of African Studies Program. University of Illinois, Urbana, 

November 1-2, 1984. 
BRUMFIT, C. and K. Johnson. 1979. Communicative approaches to language 

teaching. Oxford University Press. 
CHISHIMBA, Maurice. 1982. African varieties of English: text in 

context. Ph.D. dissertation. University of Illinois. 
COBARRUBIAS, Juan and J. A. Fishman, eds. 1983. Progress in language 

planning. The Hague: Mouton. 
CORDER, Stephen P. 1967. The significance of learner's errors. 

IRAL 5.161-170. 
CORDER, Stephen P. 1971. Idiosyncratic dialects and error analysis. 

IRAL 9.147-159. 
CROFT, Kenneth, ed. 1980. Readings in English as a second language 

for teachers and teacher trainees. Cambridge, Mass: Winthrop. 

(2nd edition) . 
DAS, Kamla. 1980. An introduction. In K. N. Daruwala, ed.: Two 

decades of Indian poetry: 1960-80. Bombay: Vikas Publishing 

DAVIDSON, David M. 1980. Current approaches to the teaching of grammar 

in ESL. In Croft,^ed., 1980: pp. 317-338. 
DECHERT, Hans W., M. Bruggemeier and D. Futterer. 1984. Transfer 

and interference in language: a select bibliography. Amsterdam 

Studies in the History and Theory of Linguistics V. Library and 

Information Sources in Linguistics, Vol. 14. Amsterdam, Philadel- 
phia : John Benjamins. 
DILLER, Karl. 1981. Natural methods of foreign language teaching: 

can they exist? What criteria must they meet? In H. Wintz, ed.: 

Native and foreign language acquisition. Annals of the New York 

Academy of Sciences 379.75-86. 
DULAY, Heidi C. and M. K. Burt. 1974. You can't learn without goofing: 

an analysis of children's second language errors. In Richards, 

ed., 1980: pp. 95-123. 
DULAY, Heidi C. and M. K. Burt. 1976. Creative construction in second 

language teaching and learning. In H. D. Brown, ed.: Papers in 

second language acquisition, pp. 65-80. Ann Arbor, MI: Univer- 
y sity of Michigan Press. 
DUSKOVA, L. 1969. On the sources of errors in foreign language learning. 

IRAL 7.11-30. 
ECKMAN, Fred. 1977. Markedness and the contrastive analysis hypothesis. 

LL 27:315-330. 


EZEKIEL, Nissim. 1976. Goodbye party for Miss Pushpa T.S. In R. 

Parthasarathy, ed.: Ten twentieth century Indian poets. Oxford 

University Press. 
FINNOCHIARO, M. and C. Brumfit. 1983. The functional -notional approach. 

Oxford University Press. 
FISHMAN, Joshua A., R. L. Cooper and A. W. Conrad, eds. 1977. 

The spread of English. Rowley, Mass.: Newbury House. 
FRIES, Charles C. 1945. Teaching and learning English as a foreign 

language. Ann Arbor, MI: University of Michigan Press. 
GAGE, William. 1961. Contrastive studies in linguistics: a biblio- 
graphical checklist. Center for Applied Linguistics, Washington, 

GONZALES, Andrew. 1983. When does an error become a feature of Philip- 
pine English? In R. B. Noss, ed.: Varieties of English in South- 

East Asia, pp. 150-172. Singapore: RELC 
HALLIDAY, M. A. K. and R. Hasan. 1976. Cohesion in English. London: 

HAMMER, John and F. A. Rice. 1965. A bibliography of contrastive 

linguistics. Center for applied linguistics, Washington, D.C. 
HANCOCK, Ian F. and R. Angogo. 1982. English in East Africa. 

In Bailey and Gorlach, eds., 1982: pp. 306-323. 
HATCH, Evelyn M., ed. 1978a. Second language acquisition: a book 

of readings. Rowley, Mass.: Newbury House. 
HATCH, Evelyn M. 1978b. Discourse analysis and second language acquis- 
ition. In Hatch, ed., 1978a: pp. 401-435. 
KACHRU, Braj B. 1976. Models of English for the Third world: white 

man's linguistic burden or language pragmatics? TESOL Quarterly 

KACHRU, Braj B., ed. 1982. The other tongue: English across cultures. 

Urbana, IL, University of Illinois Press. (Also published in 

paperback by Pergamon Press, 1983). 
KACHRU, Braj B. 1983. The Indianization of English: the English 

language in India. Oxford University Press. 
KACHRU, Braj B. 1984. The alchemy of English: the spread, functions 

and models of non-native Englishes. Pergamon Press, (in press). 
KACHRU, Yamuna. 1982. Linguistics and written discourse in particular 

languages: English and Hindi. ARAL 3.50-77. 
KACHRU, Yamuna. 1983. Cross-cultural texts, discourse strategies 

and discourse interpretation. To appear in L. Smith, ed.: Discourse 

across cultures: strategies in world Englishes. Pergamon Press. 
KACHRU, Yamuna. 1984. Discourse strategies, pragmatics and ESL: 

where are we going? To appear in RELC Journal, December, 1985. 
KUFNER, H. L. 1962. The grammatical structures of English and German. 

Chicago, University of Chicago Press. 
LADO, Robert. 1957. Linguistics across cultures. Ann Arbor, MI.: 

University of Michigan Press. 
LEE, W. R. 1968. Thoughts on contrastive linguistics in the context 

of foreign language teaching. In James E. Alatis, ed.: Contrastive 

linguistics and its pedagogical implications, Georgetown Monograph 

on languages and linguistics 21, pp. 185-194. Georgetown University, 

Washington, D.C. 
LOWENBERG, Peter. 1984a. English in the Malay Archipelago: nativization 

and its functions in a sociol inguistic area. Ph.D. dissertation. 

University of II 1 inois. 


LOWENBERG, Peter. 1984b. Non-native varieties of English: new chal- 
lenges for TESL. Presented at the Illinois TESOL/BE Central- 
Southern Regional Conference. February 4, 1984. 
MAGURA, Benjamin. 1984. Style and meaning in African English: a 

sociolinguistic study. Ph.D. dissertation. University of Illinois. 
MOULTON, W. G. 1962. The sounds of English and German. Chicago: 

University of Chicago Press. 
MUNBY, John. 1978. Communicative syllabus design. Cambridge University 

NELSON, Cecil. 1984a. Intelligibility: the case of non-native varieties 

of English. Ph.D. dissertation. University of Illinois. 
NELSON, Cecil. 1984b. My language, your culture: whose communicative 

competence? Paper presented at the joint AAAL/ACTFL Conference, 

Chicago, November 16, 1984. 
PLATT, John, H. Weber and M. L. Ho. 1984. The new Englishes. London: 

Routledge and Kegan Paul. 
RAVEM, Roar. 1968. Language acquisition in a second language environ- 
ment. IRAL 6:175-185. 
RAVEM, Roar. 1974. The development of wh- questions in first and 

second language learners. In Richards, ed., 1974: pp. 134-155. 
RICHARDS, Jack C, ed. 1974. Error analysis: perspectives on second 

language learning. London: Longman. 
RICHARDS, Jack C. 1984. The secret life of methods. TESOL Quarterly 

RICHARDS, Jack C. and Mary Tay. 1981. Norm and variability in language 

use and language learning. In Smith, ed., 1981: pp. 40-55. 
RUBIN, Joan and B. Jernudd, eds. 1975. Can language be planned? 

Honolulu: University Press of Hawaii. 
SAVIGNON, Sandra. 1983. Communicative competence: theory and classroom 

practice. Reading, Mass: Addison-Wesley. 
SCHACHTER, J. and M. Celce-Murcia. 1980. Some reservations concerning 

error analysis. In Croft, ed., 1980: pp. 120-131. 
SELINKER, Larry. 1972. Interlanguage. IRAL 10.201-231. 
SHAW, Willard D. 1981. Asian student attitudes towards English. 

In Smith, ed., 1981: pp. 108-122. 
SKUTNAAB-KANGAS, Tove. 1984. Bi 1 ingual ism or not : the education 

of minorities. Clevedon, Avon: Multilingual Matters. 
SMITH, Larry, ed. 1981. English for cross-cultural communication. 

London: Macmillan. 
SMITH, Larry, ed. 1983. Readings in English as an international language. 

Pergamon Press. 
SRIDHAR, S. N. 1980. Contrastive analysis, error analysis and inter- 
language: three phases of one goal. In Croft, ed., 1980: pp. 

STOCKWELL, Robert and J. D. Bowen. 1965. The sounds of English and 

Spanish. Chicago: University of Chicago Press. 
STOCKWELL, Robert, J. D. Bowen and J. W. Martin. 1965. The grammatical 

structures of English and Spanish. Chicago: University of Chicago 

STREVENS, Peter. 1977. New orientations in the teaching of English. 

Oxford University Press. 
STREVENS, Peter. 1981. Teaching English as an international language. 

Pergamon Press. 


WIDDOWSON, H. 1978. Teaching language as communication. Oxford 

University Press. 
WILKINS, D. 1976. Notional syllabuses: a contribution to foreign 

language curriculum. Oxford University Press. 

Studies in the Linguistic Sciences 
Volume 15, Number 1, Spring 1985 

Nkonko Mudipanu Kamwangamalu 

Perlmutter and Postal (1983) claim that the rule of 
passive can be better described in terms of grammatical 
relations, using primitives or traditional notions such as 
subject of, direct object of, and indirect object of, generally 
referred to as TERMS. On the basis of these, it is claimed that 
in relational network, only TERMS and nothing else can passivize. 
In other words, locatives and instrumentals are excluded from 
participating in passivization. The main objective of this paper 
is to show that Tshiluba and other Bantu languages (cf. Dalgish 
1976) constitute a counter-example to Perlmutter and Postal's 
(1983) claim. In particular, it will be shown here that 
passivization in Tshiluba involves locatives, in addition to TERMS. 
Data are presented which show that locatives in Tshiluba do all the 
things that TERMS do : (1) they can govern agreement on the verb, 
(2) they can passivize in both intransitive and transitive clauses. 
The question that arises here is whether these passivizing 
locatives should be treated as TERMS in Relational Grammar. This 
paper strongly argues that they should and provides, in conclusion, 
some suggestions as to how Relational Grammar could account for the 
facts presented here and elsewhere in the Bantu literature. 

1.0 Introduction 

It has been claimed in Relational Grammar (Perlmutter and Postal 
(1983) that only TERMS, that is, traditional notions such as subject of, 
direct object of and indirect of, can passivize. In other words, non-TERMS 
such as locatives, instrumentals, manners, etc. are excluded from 
passivization. However, Dalgish (1976) presents data from Olutsootsoo, a 
Bantu language of Kenya, which show that locatives in this language can 
passivize. As of this writing, no analysis has been proposed that we know 
of within Relational Grammar (hereafter RG) that accounts for Dalgish' s 
conclusions . 

The main objective of this paper is to provide evidence from another 
Bantu language, Tshiluba, spoken in the Republic of Zaire, in support of 
Dalgish' s claim, and to argue that passivization is not limited to TERMS 
only. More specifically, it will be shown that in addition to TERMS, 
Tshiluba allows the passivization of locatives in various types of clauses. 
Locatives are shown to behave like TERMS not only with regard to 
passivization, but also in terms of their ability to control subject-verb 


agreement. Moreover, Tshiluba locatives, like TERMS, can undergo rules 
such as raising, topicalization, left-dislocation, to cite some. With 
regard to advancement, an analysis will be presented which shows that 
locatives can passivize directly from their underlying position in an 
active clause to the subject position in the passive clause. This will 
be called direct passivization of the locatives. These last three 
points, that is, locative raising, topicalization/lef t-dislocation and 
direct passivization as well as constructions involving ditransitive 
verbs plus a locative were not treated in Dalgish (1976) . They will 
therefore be discussed at length in this paper for they constitute, 
in addition to agreement, strong evidence for the termhood of Bantu 
locatives . 

The paper is divided into two main parts. Part one provides, first, 
an overview of the basic morpho-syntactic characteristics of Tshiluba to 
familiarize the reader with the Tshiluba agreement system which is one 
of the crucial pieces of evidence for the arguments to be advanced in 
this paper. Second, a review of the treatment of passive in RG will be 
presented to provide the background against which the data on the 
passivization of locatives are analyzed. Part two presents a description 
of passivization in Tshiluba within the RG framework along with a 
discussion of locative raising, locative topicalization and/or left- 
dislocation and direct passivization. This section of the paper examines 
several types of passive sentences in Tshiluba, with particular emphasis 
on locatives. It shows that locatives passivize in clauses with mono- 
transitive verbs, clauses with ditransitive verbs, and in clauses with 
intransitive verbs and that they can, like TERMS in RG, undergo raising 
and left-dislocation, to cite just these two. Data from other Bantu 
languages will be referred to in order to provide a broader view of the 
phenomenon under consideration. 

The abbreviations used in this paper include the following : 
np: noun prefix; Loc: locative; T/A : tense and aspect; Ag : agreement; 
FV : final vowel; PSV : passive ; SVA : subject-verb agreement. 

2.0 Tshiluba Morpholgy and Syntax 

This section sketches Tshiluba syntax and noun and verb morphology. 
Tshiluba sentences will be translated wherever appropriate English 
equivalents are possible. Otherwise, only word-for-word glosses will be 

2,1 Morphology . Morphology, both nominal and verbal, is very important 
in Bantu syntax. Nominal morphology provides a clue to determining the 
type of agreement that must obtain between a subject noun and a verb. 


Verbal morphology determines, depending on the type of affix in presence, 
whether the affixed verb requires one or more objects. As it will be 
shown in the next section, a verbal suffix affixed to an intransitive 
verb stem, for example, may syntactically change an inherently 
intransitive verb into a transitive verb. 

2.1,1 Noun Morphology . A noun in a Bantu language in general and 
in Tshiluba in particular consists of two morphemes : a noun prefix and 
a verb stem : 

(1) a. di- tuku 'day' 

b. mu- kaji 'woman' 

c. ka- ba 'trap' 

In (1), di- , mu- , and ka- are noun prefixes; -tuku, -kaji, and -ba are 
noun stems. For each noun stem, there are generally two noun prefixes, 
one singular and the other plural.^ Thus, the plural correspondences for 
the nouns in (1) are as follows : 

(2) a. ma- tuku 'days' 

b. ba- kaji 'women' 

c. tu- ba ' traps ' 

When various basic nouns like these are analyzed in terms of singular/ 
plural correspondence, several pairings emerge, traditionally referred 
to as noun classes in Bantu linguistics (cf. Bleek 1869, Guthrie 1967, 
etc.). Bokamba (1976) defines a noun class prefix as one of the distinct 
patterns of prefix agreement that a particular language may have, with 
the actual number of noun classes for such a language being determined by 
the distinct patterns of agreement exhibited. These noun class prefixes 
are traditionally grouped in pairs such as 1/2, 3/A, 5/6, etc. as in 
Table 1 below. The range of noun classes varies from one Bantu language 
to another (cf . Guthrie 1967) . In Tshiluba, there are eighteen noun 
classes as indicated in Table 1. Three of these, class 16 pa- (on) ; 
class 17 ku-(to/at) ; and class 18 mu - (in), to which attention will 
mostly be focussed in this paper, are locatives. 


Table 1 

Tshiluba Noun Class and Subject-Verb-Agreement (SVA) System 

class no. 













u - 


person/man is . . . 




u - 


father is ... 







persons/men are... 







fathers are. . . 





u - 


tree is ... 





i - 


trees are . . . 







day is ... 





a - 


days are . . . 







door is ... 







doors are . . . 




n -zubu 

u - 


house is ... 




n -zubu 

i - 


houses are. . . 







hoe Is ... 




n- kasu 

i - 


hoes are . . . 







trap is ... 







traps are. . . 







farm/field is... 





a - 


farms/fields are... 







writing is . . . 









on the house is... 
at the house is. . . 
in the house is ... 




In this Table, column 1 indicates the noun class number ; column 2 the 
noun prefix (np) corresponding to the noun class number in column 1 ; 
column 3 presents the agreement prefix that appears on the verb in 
case a given noun class functions as subject of a clause. For example, 
if the noun dituku (day), which belongs to class 5, is used as subject 
of a clause, the agreement prefix on the verb will be di-, as shown 
in column 4. Similarly, if this noun is used in its plural form, 
matuku (days)(cf. class 6), the corresponding agreement prefix on the 
verb will be a-, as indicated in column 4. By the same token, locatives 
exhibit the same morphological and syntactic behavior as other noun 
class prefixes do. This can be seen in the underlined locative phrases 
at the bottom of Table 1. 

2.1.2 Verbal Morphology . Athough the nominal morphology seems to be 
relatively simple, Tshiluba verbal morphology is very complex. A verb 
in Bantu languages in general and in Tshiluba in particular consists 
of an infinitive prefix or an agreement prefix followed by a tense 
marker, one or more optional derivational suffixes plus an obligatory 
inflectional suffix. It has been noted in Bantu linguistics that there 
are , morphologycally, at least six derivational suffixes in Bantu 
languages : (1) the applied, (2) the causative, (3) the reciprocal, 

(4) the reversive, (5) the passive and (6) the stative (Guthrie 1962 ; 
Scotton 1967 ; Eastman 1969 ; Givon 1971a ; Bokamba 1976 and forth- 
coming). In Tshiluba, these suffixes are : (1) applied : -il- ; 

(2) causative : - ish- ; (3) reciprocal : -angan - ; (4) reversive : -ul- ; 

(5) passive : -ibw- ; (6) stative : -an- / -Ik-. 

In addition to these simple derivational suffixes, Tshiluba has 
other types of suffixes such as (7) double causative : -ishish - ; (8) 
intensive : -ulul- ; (9) locative : - ilu ; (10) pejorative : -avi . 
Each of these ten suffixes, when affixed to the verb stem, changes not 
only the verb morphology but can also change its syntax and semantics. 
The following examples are illustrative : 

(3) a. ku - kum - a 'to beat' 


b. ku - kum - il - a 'to beat for' 

c. ku - kum - ish- angan- a 'to cause to beat each other' 

In these examples, ku - is the infinitive prefix, -kum - is the verb stem, 
-il- in b^ is the applied suffix, -ish - and -angan- in £ are respectively 
causative and reciprocal suffix, -a^ is the final vowel. As it can be 
noted here, the morphology of the verb in a_ is extended with the addition 
of the applied suffix in _b and the causative and the reciprocal in £. 


While in (3)a^ the meaning of the verb is just 'to beat', the effect of 
the added suffix in (3)b^ and c is remarkable as indicated in the gloss. 
Syntactically, the extended verb in (3)b^ requires an object and a 
benefactive whereas in (3)^ , the effect of the added suffixes speaks 
for itself as evidenced in the gloss. 

2.2. Syntax . 

2.2.1 Word order . The assumed word order in a Bantu sentence is 
subject-verb-object. An adjective follows the noun it modifies. In (4) 
below, for example, baana (children) , bakudya (they ate) and bibota 
(bananas) are respectively subject, verb and object. In (5), munene 
(fat/big) is an adjective and as such, it must follow the noun it 

(4) ba - ana ba - kudya bibota 

np - children Ag - eat bananas 
(The children ate bananas.) 

(5) mu - ntu mu - nene 

np -person Ag - fat/big 
(fat/big person) 

2.2.2. Grammatical Agreement . Each noun class in Table 1 governs 
agreement on the verb or adjective with which it occurs in a given phrase 
or sentence. Traditionally, two types of agreement are distinguished in 
Bantu linguistics : nominal or adjectival agreement and verbal agreement. Adjective/Noun Agreement . This type of agreement generally 
involves the copying of the appropriate prefix of the modified noun onto 
the modifying adjectival stem as shown in (5) above as well as in the 
following additional examples. 

(6) di - tama di - kese 

np - cheek Ag - small 
(small cheek.) 

(7) mu- nzubu mu -nene 

np/Loc- house Ag - big 
(in the big house.) 


(8) ma - kasa r 

np - feet I 
(long feet) 

long Subject/Verb Agreement . Verbal agreement is characterized by 
the copying of the features specification of the subject noun or pronoun, 
which include animacy, class membership, person and number, onto the verb 
(cf. Table 1, column A.). Each verb must agree in number, noun class 
and person with its subject (there is no object-verb agreement in 
Tshiluba) by means of a prefix as can be seen in the following examples: 

(9) ku - nzubu ku - aka ku - di muntu 

np/Loc- house Ag - there Ag - is person 

at house there is a person 
(There is a person at that house.) 

(10) mu - nzubu mu- amwa mu - di muntu 

np/Loc- house Ag- there Ag - is person 

in house there is a person 
(There is a person in that house.) 

(11) pa - mesa (p)a- (p)a pa - di bintu 

np/Loc- table Ag - there Ag - are things 

on table there are things 
(There are things on that table.) 

(12) bi -bota bi - di pa -mesa 

np -bananas Ag - are np/Loc- table 

bananas are on table 
(Bananas are on the table.) 

The locatives ku-, mu-, and £a- are shown to govern agreement on the 
verb in (9), (10) and (11) respectively. In (12), the TERM bibota 
(bananas) governs agreement on the verb -di (are) by means of the 
prefix bi- . To sum up, this section has been concerned with a brief 
presentation of Tshiluba morphology and syntax . This was to introduce 
the reader to the Tshiluba agreement system and to show that all noun 
prefixes in Tshiluba, including locatives, govern agreement on the verb. 
For a more detailed discussion on agreement in Bantu languages, the 


reader Is referred to Bokamba (1976, and forthcoming). I will now 
discuss the passive in Relational Grammar. 

3.0 Passivization in Relational Grammar 

Passlvization is a central problem in RG. This theory argues that 
many attempts have unsuccessfully been made in structuralist and trans- 
formational grammar to characterize the passive in terms of word order, 
case marking and verbal morphology. These attempts have proven un- 
fruitful because not all languages have the same word order ; not all 
languages have the same case marking ; and not all languages 
distinguish between active and passive clauses in terms of verbal 
morphology. So, any theory that attempts to characterize passivization 
in terms of either of the above criteria (i.e. word order, case marking, 
verbal morphology) is but inadequate and cannot capture generalizations 
about the passive as a universal rule. 

Relational Grammar claims that the rule of passive can be better 
described in terms of grammatical relations such as subject of (hereafter 
SU) , direct object of (DO) , and indirect object of (10) . On the basis 
of these primitives, Perlmutter and Postal (1983) assume that a clause 
universally consists of a network of grammatical relations and that 
among these are SU, DO, and 10. In the light of this assumption, they 
claim that there are two universals of passivization across languages: 

(13) 'A direct object of an active clause is the (superficial) 
subject of the corresponding passive clause. 

(14) The subject of an active clause is neither the (super- 
ficial) subject nor the (superficial) direct object of 
the corresponding passive clause. So, passive is a 
(superficial) intransitive clause' (1983:9). 

On account of the above claim and assumption, Perlmutter and Postal 
state that the passive is a rule which sanctions 1-hood -' in an 
immediately successive stratum for a nominal which is a 2 of a clause 
at a stratum in which some nominal is a 1. This formulation of the 
rule of passive implies that apart from an initial 2 of an active 
clause, nothing else can become the subject of a passive clause prior 
to advancement to 2. This is counter-examplif ied by the data discussed 
in this paper which show that 3's and locatives can passivize directly 
from their initial position in an active clause to the subject position 


in the passive clause. In so doing, the passivizing ^'s or locatives do 
not affect the grammatical relation initially borne by a 2^ (or a 2^ and 
a J^ in the case of a passivizing locative). In other words, the only 
TERM that is put en chSmage by such a direct passivization is the 
initial _1 of an active clause. I will elaborate more on this with 
several examples as we proceed. 

The key notion in characterizing passivization in terms of 
grammatical relations SU, DO, and 10 is the concept of TERMHOOD . 
RG claims that there are only three TERMS , namely those primitives just 
described. According to RG, only these three TERMS , i.e. SU (or 1) , 
DO (or 2) and 10 (or 3_ (via 2 advancement)) and nothing else can 
passivize. To avoid ill-formed clauses, Perlmutter and Postal have 
formulated laws or constraints which govern the grammatical relations 
that TERMS bear to the predicate. Among these laws are the Stratal Unique - 
ness Law , ChSmeur Law , Agreement Law , to cite just three (Perlmutter and 
Postal 1983: 88-101; cf. also Frantz 1981: 71). The Stratal Uniqueness 
Law states that each TERM bears one and only one grammatical relation 
to the predicate. The ChOmeur Law says that if some nominal, Na, bears 
a given TERM relation in a given stratum S^ , and some other nominal, Mb , 
bears the same relation in the following stratum, S+1, then Na bears the 
chomeur relation in S+1. The Agreement Law states that only nominals 
bearing TERM relation in some stratum may trigger verb agreement. These 
Laws can be illustrated as follows : 

(15)a. The hunter killed two elephants 

b. Two elephants were killed by the hunter 


the hunter 

two elephants 

By virtue of the passive rule, the TERM ( two) elephants , which is a 2 
in the initial stratum CI, becomes a ]^ in stratum C2 . Similarly, the 


TERM hunter , which is a ^ on stratum CI, can no longer by virtue of the 
Stratal Uniqueness Law, bear the same relation as ( two) elephants in 
stratum C2 and by the Chomeur Law, it is put en chSmage on this stratum. 
By virtue of the Agreement Law, the TERM ( two) elephants may govern 
agreement on the verb in the final stratum C2. I will now turn to 
passivization in Tshiluba. 

4.0 Facts about the passive in T shiluba 

In this section, various cases of passivization will be analyzed. 
I will successively discuss passivization in (1) clauses with mono- 
transitive verbs ; (2) clauses with ditransitive verbs ; (3) complex 
clauses (i.e. clauses with verbal extension) ; and (4) clauses with 
locatives in order to demonstrate that locatives in Tshiluba behave 
both morphologically and syntactically like TERMS in Relational Grammar. 

4,1 Passivization of TERMS in clauses with mono transitive verbs. 

(16) a. Mwana u- aku- di 

a di- bota 

child Ag- T/A- eat -FV np- banana 
(The child ate the banana.) 

b. Di- bota di- aku- di 


np- banana Ag- T/A- eat -PSV -FV by 
(The banana was eaten by the child.) 

kudi mwana 

(17)a. Bantu ba- aku- injil - a tshi- ibi 

men Ag- T/A- close -FV np- door 
(The men closed the door.) 

b. Tshi- ibi tshi- aku- injid- ibw 

np- door Ag- T/A- close- PSV -FV by 
(The door was closed by the men.) 

kudi bantu 

The relational treatment of (16) and/or (17) is straightforward. In 
(16)a , for example, mwana (child) and d ibota (banana) will respectively 
bear the 1^ and the ^-relation to the predicate - di-a (eat). But in (16)b, 
dibota (banana) is the superficial 1^ of the passive clause whereas 
mwana (child) is, by virtue of the Stratal Uniqueness Law and the ChSmeur 
Law, put en chfimage on the final stratum. The stratal diagram for (16) 


for example is as shown here 

(he ate) 

4.2. Passivlzation of TERMS in clauses with dltransitlve verbs. 

(18) a. Mu- ntu u- aku- p 



np- man Ag- T/A- give-FV np- children np- 
(The man gave fruit to the children.) 


b.*Mu- ntu u- aku- 

bi- muma ba- ana 

np- man Ag- T/A- give-FV np- fruit np- children 
(*The man gave the children to the fruit) 

ba- aku- p - ibw- 

a bimuma kudi muntu 

np- children Ag- T/A- give- PSV-FV fruit by 
(The children were given fruit by the man.) 

d. bi- muma bi- aku - p 

-ibw- a baana 

kudi muntu 

np- fruit Ag- T/A - give -PSV-FV children by man 
(The fruit were given to the children by the man.) 

In (18)a, muntu (man), baana (children) and bimuma (fruit) are 1^, 
3 and 2 respectively. Tshiluba distinguishes between animate and 
inanimate objects and requires that when these co-occur in a clause, 
the animate come immediately after the verb as the TERM baana does in 
(18)a, preceding the TERM bimuma , which is actually a 2^ in this 
clause. The violation of this constraint on Tshiluba word order 
results in an ungrammatical structure as shown by the unacceptability 
of (18)b where the inanimate bimuma precedes the animate baana. 
Moreover, not only does the change of the word order make (18)b 


unacceptable, but it also changes the resulting meaning. The initial 3_, 
i.e. baana (children) in (18)a can passivize as in (18)c without 
affecting the grammatical relation initially borne by the 2^, i.e. 
bimuma (fruit), to the predicate -pa (give). The same is true of 2 
when it passivizes as indicated in (18) d. So, there seems to be no 
evidence for 3^ to 2^ advancement in Tshiluba. With regard to the word 
order, it will be assumed here that when animate and regardless of 
whether it is a DO or an jLO, the object of a clause comes immediately 
after the verb. Apart from this, even if we assume that in terms of 
the grammatical relation assignment, the animate and the inanimate 
objects of an active clause in Tshiluba are initially a 2_ and a 3^ 
respectively, this will not allow 2 ^° 2_ advancement because the 
resulting structure will still prove ungrammatical as shown in (18)b. 
In other words, such an advancement implies that the hypothetical 
initial 3^, i.e. bimuma (fruit), will advance to 2 and as a result, 
the hypothetical initial 2^, i.e. baa na (children), will bear the 
chfemeur relation to the predicate -pa (give) by virtue of the Stratal 
Uniqueness Law. The resulting structure in this hypothetical 
advancement is but (18)b just discussed, which violates the constraint 
referred to above on the Tshiluba word order. So, both the grammatical 
relation assignment and the Tshiluba word order seem to provide no 
evidence at all for _3 to 2^ advancement. It will then be assumed, with 
regard to the former (i.e. grammatical relation assignment), that 
an animate 10 of an active clause is a superficial 2 and that the DO 
is a superficial 3, since no advancement of jW to DO proves to be 
allowed in Tshiluba. A rather strong claim is in order here. That is, 
advancement in Tshiluba is X (i.e. 2, 3, or Loc) to 1^. Put another 
way, there is, at least for all the cases considered here, no such a 
thing as 3^ to 2^ or Loc to 2 to 2^ advancement in Tshiluba. Several 
cases will be discussed in support of this claim which holds of 
passivization of TERMS in complex clauses to which I turn now. 

4,3. Passivization of TERMS in complex clauses . It has been noted 
in the discussion of the Tshiluba verbal morphology that there are 
about ten verbal suffixes in Tshiluba. How many of these can be affixed 
at a time to a verb stem is Immaterial here. A verb can passivize 
regardless of whether it is affixed with some of these suffixes. Given 
the semantic complexity that results when more than two suffixes are 
affixed to a verb stem, the number of suffixes on the verbal stems 
will be limited to two, namely, the causative and/or the applied. 


The following examples illustrate the passivizatlon of TERMS in clauses 
with the applied suffix: 

(19)a. Mutombo u-aku - sumb - il - a bantu bilunga 

Mutombo Ag-T/A - buy -Applied -FV men potatoes 
(Mutombo bought potatoes for the men.) 

b.*Mutombo u-aku - sumb - il - a bilunga bantu 

Mutombo Ag-T/A - buy -Applied -FV potatoes men 
(Mutombo bought men for the potatoes.) 

c. Ba-ntu ba- aku- sumb- id - ibw- a bilunga kudi M. 

np-men Ag- T/A- buy - Applied - PSV-FV potatoes by M. 
(The men were bought potatoes for by Mutombo.) 

d. Bi-lunga bi-aku- sumb- id - ibw- a bantu kudi M, 

np-potatoes Ag-T/A-buy-Applied - PSV-FV men by M, 
(The potatoes were bought for the men by Mutombo.) 

The relational treatment of these structures is similar to that 
proposed for (18)a-d above. In (19)a, I assume that Mutombo, bantu 
(men) and bilunga (potatoes) are respectively 1^, 2, and 3^. An 
alternative to this could be that the TERM bantu and bilunga be 
considered as 3_ and 2^, respectively. However, this would imply that 
there is 2. to 2^ advancement which starts out of an ungrammatical 
structure listed in (19)b to end up with the structure in (19)a. This 
alternative is inconsistent with the notion of animacy/inanimacy 
discussed above, which requires that the animate object of a clause, 
regardless of whether it is a DO or an 10, come initially after the verb. 
It is worth mentioning here that there is no need to provide cases of 
passivizatlon in clauses with the causative suffix since the rule of 
passive in such clauses operates exactly as in those with the applied 
suffix. In other words, all the TERMS involved in constructions with the 
causative suffix do indeed passivize in the way already described above 
and do obey the constraint on Tshiluba word order and grammatical 
relation assignment. This constraint has already been illustrated in 
clauses such as (18)a and (19)a each of which has two objects: one 
animate and the other inanimate. But what happens in case two animate 
objects co-occur in the same clause? What is the syntactic behavior of 
TERMS In such a case? It has been demonstrated that when an animate 


object and an inanimate object co-occur in a clause, the former comes 
immediately after the verb. In case two animate objects co-occur in 
a clause, the animate 10 must, regardless of the presence of the 
animate DO, come immediately after the verb. This provides more 
evidence for the claim that there is no istance of (Loc to) 2. ^° 1 
advancement in Tshiluba but that there is, instead, X to j^ advancement. 
The following structures are illustrative: 

(20) a. Baaba u- aku- p - a taatu baana 

mother Ag- T/A- give-FV father children 
(Mother gave the children to Daddy.) 

b.*Baaba u- aku- p - a baana taatu 

mother Ag- T/A-give -FV children father 
(Mother gave Daddy to the children.) 

c. Baana ba-aku- p - ibw- a taatu kudi baaba 

children Ag-T/A-give- PSV-FV father by mother 
(The children were given to Daddy by mother.) 

d, Taatu u-aku- p - ibw - a baana kudi baaba 

father Ag-T/A- give- PSV -FV children by mother 
(Daddy was given the children by mother.) 

Here again, the relational treatment of these structures is straight- 
forward. The TERM baaba (mother), taatu (daddy) and baana (children) 
in (20)a are respectively 1^, (superficial) 2 and (superficial) 3. An 
alternative to this could be that proposed for (19) according to which 
the TERM taatu (daddy) and baana (children) could be considered to 
have undergone 2 '^o ^^ advancement starting out of the ungrammatical 
structure given in (20)b to provide the structure listed in (20)a. 
However, it has already been demonstrated that this alternative is 
incompatible with the claim that when two animate objects co-occur 
in a clause, the Tshiluba syntax requires that the 10 come, regardless 
of the presence of the animate DO, immediately after the verb. This 
explains, as a matter of fact, why the change of the word order operated 
in (20)b^ and the meaning resulting thereof are but unacceptable in 
Tshiluba. When the constraint on the Tshiluba word order is not violated, 
the resulting structures prove entirely grammatical and acceptable as 


illustrated in (20)£ and d^. The TERM baana (children), a superficial 
3 in (20)a, has passivized directly from this position to the subject 
p^osition in the passive clause (20)£. Note that in so doing, it does 
not bump taatu (daddy), the superficial 2_, into chomage. Similarly, 
the TERM taatu , a superficial 2_ in (20)a, passivizes directly from 
this position to the subject position in the passive clause (20)d^. 

Some of the cases considered so far prove to be handled perfectly 
by means of grammatical relations in terms of TERMS. Others, however, 
are left open. Contrary to the claim in RG that no 3^ advances to 1^ 
without advancing first to 2, Tshiluba Vs appear to passivize directly 
from their initial position in an active clause to the subject position 
in the passive clause. The most crucial fact here is that the 3^'s 
going to Vs do not bump existing 2^'s to 2^'s, which they should by the 
Stratal Uniqueness Law if they actually advanced to 1^' s via an 
intermediate 2-stage. This way of passivizing a TERM directly from 
whatever position in an active clause to the subject position in the 
passive clause without affecting the grammatical relations initially 
ijorne by other TERMS (or non-TERMS) has been referred to above as 
' direct passivization of TERMS ' . It will be shown hereafter that 
direct passivization of TERMS holds, not only of the Tshiluba data, 
but also of the following Lingala and Kiswahili data which show that 
3 to 2 advancement is not operative in these Bantu languages either. 


(21) a. Mama a -somb - el - i mu-ana li- kemba 

mother Ag -buy -Applied-FV np-chlld np- plantain 
(Mother bought a plantain for the child.) 

b . *Mama a -somb - el - i li-kemba mu-ana 

mother Ag -buy -Applied-FV np-plantain np-child 
(*Mother bought a child for the plantain.) 

c. Mu-ana a -somb - el - am - i likemba (na mama) 

np-child Ag -buy -Applied-PSV-FV plantain (by mother) 
(The child was bought a plantain for by mother.) 

d, Li-kemba li- somb- el - am - i mwana (na mama) 

np-plantain Ag- buy - Applied- FSV-FV child (by mother) 
(The plantain was bought for the child by mother.) 



(22)a. Mama a - li - (m) - nunu- 11 - a mtoto ndizl 
mother Ag - T/A-(OP) - buy-Applied-FV child bananas 

(Mother bought bananas for the child.) 

b.*Mama a - li -(zi) - nunu- li - a ndizi mtoto 

mother Ag - T/A-(OP) - buy -Applied-FV bananas child 
(*Mother bought the child for the bananas.) 

c, Mtoto a - li - nunu - li - w - a ndizi 

child Ag - T/A- buy -Applied- PSV -FV bananas 
(The child was bought bananas for by mother.) 

d, Ndizi zi- li - (m )- nunu - li - w - a mtoto 

bananas Ag- T/A- (OP)- buy -Applied-PSV-FV child 
(The bananas were bought for the child by mother.) 

These data indicate that the TERM mwana (child) and llkemba (plantain) 
in (21) a can passivize directly as in (21)£ and d^. Similarly, the 
TERM mtoto (child) and ndizi (bananas) in (22)a can passivize 
directly as shown in (22)£ and d respectively. Note that here again, 
none of these passivizing TERMS affects the grammatical relation 
initially borne by other TERMS to the predicate. Unless we assume 
that _3 '^° A advancement starts out of ungrammatical structures such 
as the Tshiluba (18)b, (19)b, and (20)b; Lingala (21)b and Kiswahili 
(22)b, it makes no sense to speak about 3^ to 2^ advancement or 
X to 3 to 2^ advancement in these languages. Let us now turn to 
passivizing locatives in Tshiluba. 

5.0 Tshiluba passivizing locatives . 

In this section, I will discuss the passivization of locatives 
in simple and complex sentences. It is shown that locatives passivize 
freely in transitive as well as intransitive clauses. Support fpr the 
termhood of locatives lies not only in their independently motivated 
passivizing in such clauses, but also in their syntactic property of 
governing agreement on the verb whenever they are used as clause head. 
In addition, constructions involving raising and left-dislocation 


will be considered which show that both TERMS and the so-called 
non-TERMS, i.e., locatives, can undergo object to subject raising, 
subject to subject raising, and left-dislocation. Also, it is 
demonstrated that locatives, like Vs in Tshiluba, can undergo 
what we have referred to as direct passivization. In other words, 
there is no such a thing as Loc to 3^ to 2^ to j^ advancement in this 
language. There is , instead, Loc to 1^ advancement. It will be argued 
that both 3's and locatives undergo direct passivization. That is, 
again, any passivizing locative or TERM does not affect the existing 
grammatical relation initially borne by other elements, TERMS or 
non-TERMS, to the predicate. Data from other Bantu languages will be 
presented, as above, in support of this argument. To start with, let 
us consider the passivization of locatives in clauses with mono- 
transitive verbs. 

5 .1 Passivization of locatives in clauses with mono transitive verbs . 

(23) a. Mwivi u-aku- kum - a baana mu - nzubu 

thief Ag-T/A- beat -FV children np/Loc- house 
(The thief beat the children in the house.) 

b.*Mwlvi u-aku- kum - a mu - nzubu baana 

thief Ag-T/A- beat -FV np/Loc- house children 
(*The thief beat in the house the ch-'ldren) 

c, Mu - nzubu mu - aku- kum- ibw- a baana kudi rawlvi 

np/Loc- house Ag - T/A- beat-PSV-FV children by thief 
(Lit: In the house were beaten the children by the thief.) 

d. Baana ba -aku- kum- ibw - a mu -nzubu kudi mwivi 

children Ag -T/A-beat- PSV -FV np/Loc-house by thief 
(The children were beaten in the house by the thief.) 

(23)b is ruled out by the constraint of animate/inanimate in Tshiluba, 
Nevertheless, one would expect that the locative mu-nzubu (in the 
house) in (23)a will first advance to 2 before advancing to 1^ as it 
is required in RG. However, Loc to 2_ advancement is ruled out because 
the resulting structure proves ungrammatical as indicated in (23)b. 
Instead, the locative mu-nzubu passivizes directly from its initial 
position in (23)a^ to its actual position in the passive clause (23)£. 


Other locative phrases in Tshiluba behave the same way. The following 
data from Kiswahili, also a Bantu language, attest that locatives in 
this language too do not undergo Loc to 2 advancement, if ever such 
a thing as locative advancement or promotion could fit in the 
hierarchy of TERMS in Relational Grammar. 


(24) a. Mama a - li - wek - a chakula kwenye meza 

mother Ag - T/A- put -FV food Loc (on) table 
(Mother put food on the table.) 

b.*Mama a - li - wek - a kwenye meza chakula 

mother Ag - T/A- put -FV Loc (on) table food 
(*Mother put on the table food) 

c. Kwenye meza ku - li- wek- w - a chakula (na mama) 

Loc (on) table Ag - T/A-put-PSV-FV food (by mother) 
(Lit: On the table was put food by mother.) 

i2h)h attests to the fact that locative phrases in Kiswahili, like 
the Tshiluba case discussed above (cf. (23)b) , do not undergo Loc 
to 2 advancement. The structures discussed so far strongly support 
the^claim that passivization of locatives, at .least in the languages 
considered here, is a one-step process, that is, Loc to 1^ advancement 
rather than Loc to _3 to 2^ to j^ as required in RG. This claim also 
holds of the passivization of locatives in clauses with ditransitive 
verbs as it will be shown in the following section. 

5.2 Passivization of locatives in clauses with ditransitive verbs 

(25) a. Mu-longeshi u-aku- p - a baana bibota ku -kalasa 

np- teacher Ag-T/A-give-FV children bananas Loc (at) -school 
(The teacher gave bananas to the children at school.) 

b . *Mulongeshi u-aku- p - a baana ku - kalasa bibota 

teacher Ag-T/A-give-FV children Loc- school banana 
(*The teacher gave at school the children bananas.) 


c .*Mulongeshi u-aKu- p - a ku -kalasa baana bibota 

teacher Ag-T/A-give -FV I.oc-school children bananas 
(*The teacher gave at school the children bananas.) 

d. ku -kalasa ku-aku- p -ibw- a baana bibota kudi mulongeshi 

Loc -school Ag-T/A-give-PSV- FV children bananas by teacher 
(Lit: At school were given the bananas to the children by the 
teacher. ) 

In (25)a, Mulongeshi (teacher), baana (children), bibota (bananas) and 
ku-kalasa (at school) are 1, 3 (or superficial 2), 1_ (or superficial 3^) 
and locative respectively. The ungrammaticality of (25)b and £ indicates 
that there is no Loc to 2. to 2^ advancement in Tshiluba. Instead, the 
locative ku-kalasa (at school) is shown to passivize directly from its 
original position in (25)a^ to its actual position in the passive clause 
(25)d^. The crucial point here is that in passivizing directly as above, 
this locative does not affect the initial 2 and 2_. These two remain the 
same, bearing their respective original grammatical relation to the 
predicate - pa (give) . 

To sum up, RG proves empirically inadequate because, contrary to 
the claim it makes that only TERMS can passivize, it provides no way to 
account for cases like those discussed here in which it has been 
sufficiently demonstrated that locatives, like TERMS, can indeed 
passivize freely in all kinds of constructions, namely, constructions 
with monotransitive verbs , constructions with ditransitive verbs, and 
those with verbal extension. Moreover, it will be shown in the following 
section that locatives can passivize in intransitive clauses as well, 
i.e., independently of whether they occur with some nominal in a clause. 
In other words, I consider locatives as object in the way described in 
Dalgish (1976) and Stucky (1976) and as such, they can transitivize 
intransitive clauses and passivize in the latter as illustrated in (26) ; 

(26)a. Mwana u-di mu- iman- a pa -mesa 

child Ag-is Ag-stand-FV Loc(on)-table 
(The child is standing on the table.) 

b. Pa -mesa pa-di pa- iman-ibw- a kudi mwana 

Loc (on) -table Ag-is Ag-stand-PSV-FV by child 

(Lit: On the table is stood by the child.) 
The table is stood on by the child 

(26)_b is the passive counterpart of the active intransitive clause (26)a. 
Note also the passive suffix -ibw- on the verb - iman - (stand) . Here again. 


RG has provided no analysis that is known of which would account for 
sentences like (26)b^ in which the locative pa-mesa (on the table) is 
subject. Before drawing a general conclusion, the rule of raising and 
left-dislocation will be discussed in order to examine the morphological 
and syntactic behavior of TERMS and locatives in constructions involving 
these rules. It will be argued here that these rules can apply without any 
modification whatsoever to both TERMS and locatives. 

5.3 Locative Raising and Left-Dislocation in Tshiluba . Constructions 
which involve raising in Tshiluba require the verb ku-mwenek-a (seem) or 
the phrase bu mutu ne (as if) either of which can be used with any of the 
following forms of the verb ku-ikal-a (be) to yield phrases similar to the 
English you seem , he seems , it seems , etc. (27)a-c illustrates some of the 
forms of the irregular verb ku-ikala (be) whereas (28) shows the 
combination of these forms with the verb kumweneka (to seem) to yield the 
equivalent of the English phrases listed in the gloss. 

(27) a. u-di/ u'-vwa'' (You are/were) 

b. u-di7 u-vw^ ((s)he is/was) 

c. bi-di/ bi-vwa" (it is / was) 

(28) a. u-di/ u-vwa u - mwenek - a (ne) 

you-are/you-were Ag-seem- FV (that) 
(You seem/seemed (that).) 

b. u-di /u-vwa u - mwenek - a (ne) 

he-is /he-was Ag- seem - FV (that) 
(He/she seems/seemed (that) .) 

N. ^ •. / 

c. bi-di/ bi-vwa bi- mwenek - a (ne) 

it-is/ it-was Ag- seem -FV (that) 
(It seems/seemed (that).) 

bi in (28)£ la a neutral prefix generally used for unspecified subject. 
The phrase ~bi-di bimwenek-a (it seems) in (28)£ can be used in 
constructions such as (29)a below in which object to subject raising 
(hereafter OSR) and subject to subject raising (SSR) may apply to yield 
the structures given in (29)b-f. 

(29) a. Bi-di bi-mwenek-a ne baana ba-vwa ba-tek- a 

it-is Ag-seem-FV that children Ag-were Ag-put-FV 

bu-kula pa - makala 
np-four Loc(on)- charcoal 


(It seems that the children put the flour on the charcoal.) 

b. Ba-ana ba-di ba-mwenek-a ne ba-vwa ba-tek-a bu-kula 
np-children Ag-are Ag-seem-FV that Ag-were Ag-put-FV np-flour 

pa - makala 
Loc(on)- charcoal 

(The children seem that they put the flour on the charcoal.) 

c. Bu-kula bu-di bu-mwenek-a ne baana ba-vwa ba-bu-tek- a 
np-flour Ag-is Ag-seem- FV that children Ag-were Ag-OP-put-Fy 

pa - makala. 
Loc (on) - charcoal . 
(The flour seems that the children put i_t on the charcoal.) 

d.*Bu-kula bu-di bu-mwenek-a ne baana ba-vwa ba-teka 
np-flour Ag-is Ag-seem- FV that children Ag-were Ag-put 

pa - makala 
Loc (on) - charcoal. 
(*The flour seems that the children put on the charcoal.) 

e. Pa -makala pa-di pa-mwenek- a ne baana ba-vwa 
Loc (on) -charcoal Ag-is Ag-seem -FV that children Ag-were 

ba-pa-tek- a bukula 

Ag-OP-put-FV flour 

(Lit: On the charcoal seems that the children put on ^t the 

f.*Pa -makala pa-di pa-mwenek- a ne baana ba-vwa 
Loc(on)-charcoal Ag-is Ag-seem -FV that children Ag-were 

ba-tek- a bukula. 
Ag-put-FV flour . 
(On the charcoal seems that the children put the flour.) 


Here, (29)_b is the case of subject to subject raising (SSR) . In this 
structure, the TERM baana (children) .which is actually the subject of 
the embedded clause baana bavwa bateka bukula pamakala (the children put 
the flour on the charcoal), is raised to the subject position In the matrix 
clause. The difference between (29)^ and b^ is that in the former, the 
dummy bi-di (it is) governs agreement on the verb -mweneka (seem) whereas 
in the latter, it is not the dummy but rather the TERM baana (children) 
that governs agreement all the way through by means of the prefix ba- (cf . 
(29)b ). (29)£ is the case of object to subject raising (OSR) . The TERM 
bukuTa (flour) .which is actually the object of the embedded clause referred 
to above, has been raised to the subject position in the matrix clause as 
shown in (29)£. Similarly, the locative phrase p a-makala (on the charcoal) 
has been raise^d to the subject position as illustrated in (29)d^. Note here 
that raising applies in the same way for both TERMS and locatives. When 
OSR applies for example, Tshiluba syntax requires that a clitic pronoun 
referring to the raised object or locative appear on the verb as indicated 
by the underlined clitic or object pronoun (OP) -bu-, referring to bukula 
(flour) in (29)£ , and the clitic -pa.-, referring to the locative pa-makala 
(on the charcoal) in i29)e . In case the clitic pronoun does not surface 
where the Tshiluba syntax requires it to, the resulting structure will be 
unacceptable as shown in (29)jd and f_. This constraint does not seem to be 
operative in the case of SSR as indicated by the grammaticality of (29)b 
in which no clitic pronoun is required to appear. Similarly, when a 
locative is the subject of an embedded clause, this constraint does not 
seem to be operative either. This point is illustrated in (30) a_ below to 
which SSR has applied to yield the structure listed in (30)b^. 

(30) a. Bi-di bi-nwenek-a ne mu -mulangi mu-di mayi 
it-is Ag-seem -FV that Loc(in)-bottle Ag-is water 
(It seems that there is water in the bottle.) 

b. Mu -mulangi mu-di mu-rawenek-a ne mu-di maayi 
Loc (in) -bottle Ag-is Ag-seem -FV that Ag-is water 
(Lit: In the bottle seems that there is water.) 

The locative mu-mulangi (in the bottle), which is initially the subject 
of the embedded clause mu-mulangi mudi maayi (in the bottle is water) in 
(30)^, has been raised to the subject position in the matrix clause as 
illustrated in (30)b. For both the TERM baana (children) in (29)b and the 
locative mu-mulangi (in the bottle) in (30)b^, the operation of SSR does 
not require that a clitic pronoun appear on the verb of the embedded 
clause. The data presented here demonstrate that locatives behave 
like TERMS not only with regard to passivization, but also, with regard to 
other rules such as raising. Apparently, this rule seems to operate like 
left-dislocation. However, although both raising and left-dislocation 


appear to share some syntactic properties as it will be shown below, these 
two rules are indeed distinct. In the case of raising, whether OSR or SSR, 
it is the raised TERM or locative that governs agreement on the verb. But 
in the case of left-dislocation, it is not the left-dislocated TERM or 
locative but rather the logical subject or locative of a clause 
that governs agreement on the verb. The following structures are 
illustrative : 

(31) a. Ba- ana ba-di ba-nanga bi-muma 

np-children Ag-are Ag-like np-fruit 
(Children like fruit.) 

b. Bi-muma, ba-ana ba-di ba-bi^-nanga 

np-fruit np-chlldren Ag-are Ag-OT^-like 
(Fruit, children like them.) 

c.*Bi-muma, ba-ana ba-di ba-nanga 

np-fruit np-children Ag-are Ag-like 
(Fruit, children like) 

C32)a. Ba-ana ba-aku-komb- a mu -nzubu 

np-children Ag-T/A-clean-FV Loc(in) -house 
(The children cleaned the house.) 

b. Mu -nzubu, ba-ana ba-aku-mu-komb- a 

Loc(in) -house np-children Ag-T/A-OP-clean- FV 
(Lit: In the house, the children cleaned in it .) 

c.*Mu -nzubu, ba-ana ba-aku-komb- a 

Loc(in) -house np-children Ag-T/A-clean-FV 
(In the house, the children cleaned.) 

Here, the TERM blmuma (fruit) is a 2^ in (31)a and so is the Loc mu-nzubu 
(in the house) in (32)a. They each have been left-dislocated as illustrated 
in (31)_b and (32)b^ respectively. In neither case, the left-dislocated 
TERM or Loc governs agreement on the verb. Left-dislocation, like OSR, 
requires that a clitic pronoun show up on the verb of the clause in which 
it applies. Otherwise, the resulting structure is but unacceptable as 
shown in (31)£ and (32)£. 

6.0 Conclusions and implications . The basic claim throughout this paper 
has been that Tshiluba locatives passivize and as such, they should be 
included in the hierarchy of TERMS in Relational Grammar. Data from other 


Bantu languages such as Olutsootsoo (cf. Dalglsh 1976), Lingala and 
Kiswahili appear to support this claim. Syntactically and morphologically, 
Tshiluba locatives do all the things that TERMS do : (1) they can govern 
agreement on the verb , which gives them support by the RG Agreement 
Law ; (2) they can undergo all the rules that apply to TERMS, amongst 
others, passive, raising, left-dislocation, to cite some ; (3) they can 
passivize in both transitive and intransitive clauses. With regard to 
advancement, the data from the languages considered here have 
sufficiently demonstrated that 3^ to _2 advancement or Loc to 3^ to 2^ to 1^ 
advancement is a questionable matter in these languages. If such an 
advancement were operative in these languages, it would have indicated 
that due to the Stratal Uniqueness Law, an initial 2 for example loses 
its 2-relation to the predicate when a 3^ passivizes. In other words, 
for a 3 to passivize, it must, according to RG, advance first to 2_ before 
advancing to 1^ and consequently, the initial 2 must be put en chomage 
and bear the chSmeur-relation to the predicate. However, this does not 
seem to be the case at least for the languages considered here. The 
direct passivization of TERMS or Locatives explains why this is not so. 
The facts discussed here show that there is an imperious need in RG to 
modify the scope of the notion of TERM and the proposed Laws to accomodate 
the data presented in this paper and elsewhere in the Bantu literature. 


I am grateful to Jerry Morgan, Georgia M. Green, Chuck Kisseberth, 
Eyamba Bokamba, Janice Jake and James Yoon for their helpful comments and 
suggestions. The remaining weaknesses and mistakes are mine. 

An earlier version of this paper was presented at the 16th Annual Conference 
on African Linguistics at Yale University in March 1985. 

Although there are generally two noun prefixes, one singular and the 
other plural, some nouns, however, do not show a singular/plural alternation, 
e.g. (a) nzubu u-di mu-nene 

house Ag-is Ag-big ( The house is small) 

(b) nzubu i- di mi-nene 

house Ag-are Ag-big (The houses are big) 

It is only through the subject-verb agreement that one can tell that nzubu 
(house) in (a) is singular and that in (b) , it is in plural (cf. Table 1 for 
SVA prefix in Tshiluba) . 

Generally speaking, Tshiluba phonology requires that 1 s become n 

in intervocalic position, V-^ V2, when V-^ is preceded by a nasal. Thus, 

ku-kum-il^-a (to beat for ) in (3)b surfaces as ku-kum-in-a. 

In RG, it is customary to use numbers such as 1, 2 and 3 respectively 
to refer to SUBJECT, DIRECT OBJECT and INDIRECT OBJECT. Similarly, 1-hood , 


2-hood and 3-hood refer, by analogy, to subject, direct object and indirect 
object respectively. 


BOKAMBA, Eyamba G. 1976. On the syntax and semantics of derivational 

verb suffixes in Bantu languages. 1975 Mid-America Linguistics 

Conference papers, pp. 38-50. 

. 1976. Question formation in some Bantu languages. Indiana University 

doctoral dissertation. Bloomington : Indiana. 

. forthcoming. Aspects of Bantu syntax. 
DALGISH, Gerard M. 1976a. Locative MPs, locative suffixes and grammatical 

relations. Proceedings of the second meeting of the Berkeley Linguistic 

Society, pp. 139-148. 
^— — . Passivizing locatives in Olutsootsoo. Studies in the Linguistic 

Sciences, 6:1.57-68 
FRANTZ, Donald G. 1981. Grammatical relations in universal grammar. 

Monograph, the Summer Institute of Linguistics : University of 

PERLMUTTER, David and P. Postal, 1974. Relational grammar : LSA linguistic 

notes. Mimeographed. 
. 1974. Some general laws of grammar. Handout, Linguistic Institute. 

. 1983. Toward a universal characterization of passivization. In 

David M. Perlmutter, ed.: Studies in Relational Grammar 1, pp. 3-29. 

Chicago : The University of Chicago Press. 
STUCKY, Susan U. 1978. Locatives as objects in Tshiluba : a function of 

transitivity. Studies in the Linguistic Sciences, 6:2.174-202. 
TRITHART, L. 1979. Topicality : an alternative to the relational view 

of Bantu passive. Studies in African Linguistics, 10:1.1-29. 

studies in the Linguistic Sciences 
Voiuwe 15, Nui.iDer 1, Spring 1965 


Tsuneko Nakazawa 

A Japanese relative clause and a noun toKi ^time', which 
syntactically behaves as the head noun of ttie relative clause, 
constitute a subordinate clause, in effect equivalent to the 
Englisli " when'-ciause. In toki -clause, however, the verb forms, 
non-past ru-form and past ta-form, do not directly reflect the 
time that the action referred to by the verb occurred. The 
present paper describes the determination of verb forms in toki - 
clauses in relation to the verb form in the matrix clause. Both 
the concepts of "aspect" and "tense", as well as subcategorization 
of verbs, are necessary to clarify the determination of the 
subordinate verb forms. 

Different authors have asserted repeatedly that the two verb 
forms are interchangeable in toki -clauscs without resulting in 
different semantic readings if the verb is a stative verb. 
However, it will be shown that the verb forms are not always 
interchangeable even if the verb is a stative verb, while the 
interchange of non-stative verb forms is also possible if certain 
conditions are satisfied. 

The present paper claims that botii the ru-form and the 
ta-form can be used in the toki -clauoC when (1) the entire 
sentence indicates a past event, and (2) the action (or state) 
indicated by the verb in the toki -clause takes place at the same 
time as that of the verb in the matrix clause. That is, when the 
actions indicated by the toki -clause and the matrix clause take 
place simultaneously, both the ru-form as an indication of 
incompletive aspect and the ta-form as an indication of past tense 
are possible in the toki -clause, regardless of whether the verb is 
stative or not. 



The term "verb form" is used in this paper simply to mean either 
the ru-form or the ta-form of a verb and does not presuppose any 
one-to-one correspondence between the verb form and the present/past 
tense indicated by the verb. 


"Tense relates the time of the situation referred to to some other 
time, usually to the moment of speaking." (Comrie 1976:1) That is, 
tense denotes a temporal relation between the time referred by tne verb 
and the time when the utterence is made. 


In a simplex sentence, tne ta-form indicate j a time prior to the 
speech act, i.e. past tense or L+pastJ, while the ru-form indicates a 
time simultaneous with or later than the speech act, present/future 
tense or [-past]. 

In a complex sentence such as a sentence with a toki -clause 
embedded, the tense of th« entire sentence is determined by the verb 
form in the matrix clause. 

(1) Taro wa daigaku ni hair-ru toki kuruiaa o Kaw-ta. 

college enter when car bought 
When Taro was about to enter college, he bought a car. 

(2) Taro wa daigaku ni hair-ta toki kuruma o kaw-ta. 

When Taio entered college, he bought a car. 

In (1) and (2), regardless of the ru - or the ta-form of the embedded 
verbs, the sentences indicate past events, L+past], due to the ta-form 
of the matrix verb, and both entering college and buying a car take 
place prior to the speech act. 

On the other hand, (3) and (4) indicate future events, [-pastj, 
and at the point of speech Taro had neither entered college nor bought 
a car . 

(3) Taro wa daigaku ni hair-ru toki kuruma o kaw-ru. 

When Taro is about to enter college, he will buy a car. 

(H) Taro wa daigaku ni hair-ta toki kuruma o kaw-ru. 

When Taro enters college, he will buy a car. 


In examples (1)-(U), the ru/ta-forms of the embedded verbs do not 
indicate tense in relation to the speech act. They show the aspectual 
relation with the time referred to by the matrix verb. 

The ru-form of a verb in a toki -clause indicates that the action 
of the embedded verb is incomplete, or [-complete], at the time of the 
action referred to by the matrix verb as in (1) and (3)i while the 
ta-form indicates that the action of the embedded verb takes place 
prior to, and is complete, or [+complete], at the time oi the action 
referred to by the matrix verb as in (2) and (U). That is, in 
toki -clause the temporal reference point of the verb generally shifts 
from the time of the speech act to the time referred to by the matrix 




In Che literature to date, casea like the above examples (l)-(4) 
are the leat.t problematic. When the ru-form is used in a toki -clause , 
tlie action ol the embedded verb (EV) has to take place after the action 
of the matrix verb (MV), and tiie actiotj referred to by the ta-form in a 
toki -clause has to take place before the action of the matrix verb. 
Good examples oi this are given by Josephs, cited by Kuno. (1973:266) 

(5) John wa basu o orx-ru toki niwa itumo tyuuj.-su-ta. EV<-MV 

bus get-off always careful-was 
When John was about to get off a bus, he was always careful. 

(6) John wa basu o ori-ta toki niwa itumo tyuui-su-ta. EV->MV 

When John had gotten off a bus, he was always careful. 

(7) John wa basu o ori-ru toki niwa itumo tyuui-su-ru. EV<-MV 

caref ul-is 
When John is about to get off a bus, he is alwayj careful. 

(8) John wa basu o ori-ta toki niwa itumo tyuui-su-ru. EV->MV 

When John has gotten off a bus, he is always careful. 

The arrow between EV and MV indicates the temporal relation between tne 
times referred to by EV and MV by pointing to the verb that reters to 
the later action. 

Many authors claim that when actions of EV and MV are considerea 
to take place simultaneously, the ta-form must be used for EV while the 
ru-form cannot be used. Soga (1983:72), for example, gives the 
following sentence as an example of simultaneous actions: 

(9) Kinou sakana o kaw-ta toki depaato e ik-ta. EV=MV 
yesterday fish bought department 

When I bought fish yesterday, I went to a department store (lit.) 

The equal sign between EV and MV indicateo the simultaneous occurrence 
of actions referred to by EV and MV. 

Kuno (1973:268-269) also makes the same conclusion that the 
ta-form of EV "represents a past action simultaneous with, or prior to, 
the action represerited by the main clause predicate", while the ru-form 
indicates only "future with respect to the time referred to by the main 
clause predicate", without giving a specific example for "simultaneous" 
case. I personally do not see how Soga's example (9) can represent the 
simultaneous occurrence of EV and MV, but in eitner case this 
generalization is proved to be lalse in the following sections. 



Tiieie are, however, wore problematic case^ where both the ru-fonn 
and ta-form in a toki -clauce are used "interchangeably" without 
changing semantic readings. Soga (1983:71) gives examples: 

(10) Kyonen Yokonama ni i-ru toki Tanaka-san ni aw-ta. EV=MV 
last year am met 

Wneci I waj in YoKohama last year, I met Mr. Tanaka. 

(11) Kyonen Yokohama ni i-ta toki Tanaka-san ni aw-ta. tV=MV 

VJhen I was in Yokohama last year, I met Mr. Tanaka. 

Soga (1!383) and Kuno (1973) (though Kuno is dealing with a relative 
clause in general) attributes tnis interchangeauility to the semantic 
subcategorization of verbs: i.e. the verb forms are interchangeable in 
case of stative verbs but they are not interchangeable in case of 
action verbs. Verbs, or more generally "verbals" to include adjectives 
arid copulas, witli a semantic feature [+stative] inuicate a state and 
what is referred to by those verbs has some duration of time, while 
[-stative] verbs are so-called action verbs atid indicate actions or 
events. »Jhen EV in a toki-clause is [+stative], the ru-form and 
ta-form can be interchanged. 

However, this interchangeability is canceled when MV nas the 

(12) Waka-j^ toki niwa okane ga na-katta. EV=MV 
young-am money had-not 

VJhen 1 was young, I did not have money. 

(13) Waka- kattd toki niwa okane ga na-kattd. EV=MV 

When 1 was young, I did not have money. 

(14) Waka-i toki niwa okane ga na-i mono des-u. EV=MV 

~ usual-is 

When people are young, they usually do not have money. 

( 15)*Waka-katta toki niwa okane ga na-i mono des-u. 

When people are young, they usually do not have money. 

In the above examples, the ru - and the ta-forms of EV in (12) and (13) 
are interchangeable but, with the ru-form for MV, the ru-form cannot 
interchanged with the ta-form and hence (15) is ungrammatical . 

To summarize the existing literature, the following rules can be 


If i£V is [-stative], EV with ru-fonn iiidicdteb EV<-MV 1,3,5,7 

ta-forin EV->MV or 2,4,6,8 

EV=MV. 9 

If EV is L+stative] and if MV is [-past], 

EV with ru-fonn indicates EV=MV. It 

If EV is [+stative] and if MV is [+past], 

EV with ru-form indicateo EV=MV 10,12 
ta-form indicates EV:MV. 11,13 

These rules, in spite ol their complexity, do not tell what will happen 
when [+3tativeJ EV takes the ta-form with L-past] MV, or if such a case 
is possible. 



To account for the interchangeability of the ru-form ana the 
ta-form of L+stative] EV, Soga (1983:73) gives an explanation: 

One reason for this phenanenon may oe that the stative verbs are 
aspectually "imperfective" and "incompletive" regardless of their 
tense forms. Since their ru-form is also imperfective and 
incompletive, there is no aspectual distinction between their ta - 
and ru-forms. 

If this were true, not all of the behavior of stative verbs would be 
accounted for. 

First, if the ru-form and the ta-form of [+stativej EV were 
equally [-complete], and if that were the reason for their 
interchangeability, then why is the interchange not possible when MV 
takes the ru-form? 

Second, there are cases where [+stative] EV does not indicate a 
simultaneous state with MV as in (16) and, not surprisingly, in such 
cases the interchange between the ru - and the ta-forms is not possible, 

(16) Gobi-sabaku dewa hiru atu- katta toki niwa EV->MV 
the Gobi daytime hot-was 
When it had been hot in the daytime at tiie Gobi, 

yoku hosi ga matataite inieta. 
often stars twinkling viewed-were 
the stars were often twinkling 


( 17)*GoDi-sabaku dewa hiru atu-_i toki niwa 
Wlien it was not in the uaytime at the GoDi, 

yoku hobi tja matataite mieta. 

the stars were often twinkling. 

(17) is ungrammaticai unless what is meant is "The stars can be seen in 
tlie daytiiiie." 

Third, there are cases where [+stative] EV takes the ta-form and 
MV takes the ru-fonn. 

(18) Byouki no toki ni undou-su-ite wa ikenaj.. EV=M\/ 
sick-are exercise-do must-not 

When you are sick, you must not do exercioe. 

(19) Byouki datta toki ni sugu undou o hazimete wa ikenai .EV->MV 
sick-were right-away start 

When you have been sick, you must not start doing exercise 
right away. 

As shown in (18) and (19), the replacement of the EV forms results in 
different meanings. 

The interchangeability between the ru-form and the ta-form of 
[+stative] EV is not because of the stativeness of EV but rather the 
simultaneous occurrence of the state referred to by EV and the action 
of MV. Therefore, even if EV is [+stative], it cannot take the ru-form 
to indicate the state completed prior to the action of MV as shown in 
(16) and (17). (18) and (19) show that the interchangeability should 
also be attributed to the ta-form of MV, which indicates that the 
entire sentence refers to the past event. 


According to the rules in Section II, the ru-form and the ta-form 
of [-stative] EV always result in a different temporal sequence of EV 
and MV. However, there are cases where the ru-form and the ta-form can 
be interchanged equally indicating ttie simultaneous occurrence of 
actions of EV and MV. 

(20) Kinou peipaa no zyunbi o su-ru toki EV=MV 
yesterday paper preparation do 
When I prepared for the paper yesterday, 

koohii o nom-ta. 
coffee drank 
I drank coffee. 


(2 1) Kiiiou peipaa no zyunbi o su-ta toKi koohii o nom-ta. EV=MV 

did EV->MV 

Wiieri I prepared for the paper yesterday, I drank cofiee. 

(22) Nihoti-ryouri o tabe-ru toki hasi o tukaw-ta. EV=MV 
Japanese food eat chopsticks used 

irfhen I ate Japanese food, I used chopsticks. 

(23) Nihori-ryouri o tabe-ta toki hasi o tukaw-ta. EVrMV 

ate EV->MV 

When I ate Japanese food, I used chopsticks. 

Evidently, the interchange between the ru-form and ta-forni is possible 
to express the temporal sequence EV=MV even if the EV is [-stative], 
though (21) and (23) are ambiguous. (21) and (23) could also mean 
EV->M\/ as the result of the completive aspect of ta-form. 

In tnese examples, zyunbi o su-ru and tabe-ru are action verbs 
and, at the same time, "durative verbs". 

Semanticaliy, durative verbs (keizoku dousi ) express the meaning of 
actions, events or processes which are perceived to require a certain 
amount of appreciable time from the inception to the termination. On 
the other hand, punctual verbs ( syunkan dousi ) are conceived not to 
requite such a time. (Soga 1983:107, parentheses added) 

This semantic feature of verbs, [+durative], is a sub-subcategory of 
L-stativeJ verbs. Therefore, verbs are either [+stative] or [-stative] 
and, furthermore, [-stative] verbs are either [+durative] or 
[-durative] . 

In case EV is [-stative, +durative], the interchange of the 
ru-form and the ta-form is possible. If MV takes the ru-form, however, 
the ru-form and the ta-form of [-stative, +durative] EV result in 
different meanings as is the case of [+stative] EV with [-past] MV. 

(24) Peipaa no zyunbi o su-ru toki wa itumo koohii o nom-ru. EV=MV 

When I prepare for papers, I always drink coffee. 

(25) Peipaa no zyunbi o su-ta toki wa itumo koohii o nom-ru. EV->MV 
When I have prepared for papers, I always drink coffee. 


To account for all the counterexamples in the previous section, 
the rules given in Section II must be modified as follows: 


If EV is [-stative , EV with ru-form indicates EV<-MV 1,3,5,7 

EV with ta-form indicates EV->MV. 2,4,6,8 

II" EV is C-stative and if MV is [-past], 

EV with ru-form indicates EV<-MV or 24 

EV=MV 24 

EV with ta-form indicates. EV->MV. 25 

If EV is [-stative and if MV is [+past], 

EV with ru-form indicates EV<-MV or 20 

EV=MV 20,22 

EV with ta-form indicates EV->MV or 21 

EVzMV. 21,23 

If EV is [+stative] and if MV is [-past], 

EV with ru-form indicates EV=MV 14,18 
EV with ta-form indicates EV->MV. 19 

If EV is [+stative] and if MV is [+past], 

EV with ru-form indicates EV=MV 10,12 

EV with ta-form indicates EV->MV or 16 

EV=MV. 11,13 

These rules look even more incomprehensible than the previous set of 
rules. To clarify the relation between the EV forms and the temporal 
sequence of MV and EV, see the following diagram, where an interesting 
generalization emerges: When and only when, the events expressed by MV 
and EV overlap on the temporal scale, EV has an option of taking either 
the ru - or ta-form regardless of whether it is [+stative] or 

Tense & Aspect of EV Form of EV 

cased) -stative +past +durative M — — S > ta-form 


case(2) -stative +past +durative M S > ta-form or 

-complete E ru-form 

case(3) -stative +past -durative 

case(4) -stative -past +durative -5 M > ta-form 


case(5) -stative -past +durative -S M > ru-form 







-S M- 



cdse(6) -stative -past -durative -S M > ru-form 

-complete E 

case(7) +stative +past M — — S > ta-form 

-••complete EEE 

case(8) +stative -t-past M— — S > ta-form or 

-complete EEE ru-form 

case(9) +3tative -past -S M > ta-form 

-•■complete EEE 

case( 10)-^-stative -past -S M > ru-form 

-complete EEE 

-> flow of time 

S point of speech act 

M event referred to by MV 

E event referred to by EV 

What happens then, is a shift of a speaker's viewpoint. Ttie tense 
[•fpast] of a verb requires its reference point to be at the time of the 
speech act, while the aspect [+complete] requires the reference point 
to be at the time referred to by MV. When MV and EV overlap on the 
temporal scale, as in cases (2) and (8), EV can take either the 
temporal reference point [-fpast], in which case EV takes the ta-form, 
or the aspectual reference point [-complete], in which case, the 
ru-form. This also explains why the option of the ta-form for EV is 
not available in case (9) although MV and EV overlap on the temporal 
scale. This is because both the temporal reference point [-past] and 
the aspectual reference point [-complete] requires the EV form to have 
the ru-form. 

The diagram also explains why examples (21) and (23) are ambiguous, 
[-stative] EV with the ta-form in (21) and (23) can represent either 
case (1) or (2). In summary, the general rules for determining the verb 
form in a toki-ciause are: 

Rule 1. If what is referred to by EV takes place simultaneously 

with what is referred to by MV, that is, EV is [-complete], 
and if the entire sentence refers to a past event, that is, 
MV takes the ta-form, then EV can take either the ru-form 
or the ta-form. (When EV is a [-durative] verb, its 
semantic property inherently does not allow the EV to 
express a simultaneous event with what is referred by MV.) 

Rule 2. Otherwise, what is referred to by EV takes place after that 
of MV, that is, EV is [-complete], then EV takes the 

Rule 3. Otherwise, EV takes the ta-form. 


Tfiese rules point out an important generalization which the previous 
rules in Section II lack. That is, the EV forms are deterinined in 
terms of the temporal sequence of actions referred to by EV and MV, and 
not in terms of a particular kind of EV. Even if EV is a stative verb, 
the interchantje of the two forms is not possible if the state referred 
to is [+complete] at the point of action of MV, while even if EV is an 
action verb, it is possiole to interchange the verb forms if the action 
of EV is [-complete] at the point of action of MV. 



So far, the phrase "interchangeable without changing the meaning" 
has been Uiied in a rough sense, but strictly all this means is that 
"the replacement of verb forms does not change the temporal sequence of 
EV and MV". It would be much more reasonable to assume that there are 
some semantic differences between [+past, -complete] EV with the 
ru-form and with the ta-form. 

The difference between the ru-form and the ta-form is in the 
temporal reference point. In (26) the reference point is the time when 
the book was taken away and in (27) the reference point is when the 
sentence was uttered. Therefore, (26) carries Taro's viewpoint, and 
possibly a subjective connotation, while (27) carries the speaker's 
viewpoint, and a more objective connotation. 

(26) Taro wa hon o yonde-i-ru toki EV=MV 

book reading-is 
When Taro was reading a book, 

sore o Hanako ni toriage-rare-ta. 
it taken-away-was 

he had it taken away by Hanako. 

(27) Taro wa hon o yonde-i-ta toki EV=MV 

When Taro was reading a book, 

sore o Hanako ni toriage-rare-ta. 

he had it taken away by Hanako. 

Since the interchange of the verb forms is a result of a shift of a 
speaker's view point, the above observation is not surprising. The 
ru-form shows the view point of the subject of the sentence and 
describes the event as it is taking place, while the ta-form shows that 
of the speaker of the sentence and describes the past event. 

Another difference is that the ta-form of a stative verb carries 
the connotation that the state no longer exists at the time of the 


speecii act. This, of couroe, is a result of trie completive aspect of 
tlie ta-fotm. 

(^6) Yainadd wa isya o site-i-ru toki kimyouna kanzya o mi-ta. EVrMV 
doctor doing-is strange patient saw 
Wlien Yamada was a doctor, he saw a strange patient. 

(.23) Yamada wa isya o site-i-ta toki kimyouna kanzya o mi-ta. EV=MV 
When Yamadd was a doctor, he saw a strange patient. 

It is natural that (29) has a stronger connotation that Yamada was not 
a doctor any longer when the sentence was uttered since the ta-form of 
EV is chosen to state the toki -clause as something [+past] while the 
ru-form emphsizes the toki -clause as [-complete] in (28). 

It follows from the above generalization that, when a toki -clause 
includes an adverb which refers to past time, the ta-form is used more 
often since such an adverb shares the feature [+past] rether than 
[-complete]. This phenomenon was first stated by Joseplis (1971) as an 
obligatory choice of a verb form in a relative clause in general, but 
as far as a toki -clause is concerned, the ta-form with such an adverb 
is not obligatory. 

(30) Taro wa Hanako ga kinou benkyou site-i-ru toKi EV=MV 
yesterday study doing-is 
When Hanako was Studying yesterday, 

sore o soba de mite-i-ta. 
it nearoy watching-was 
Taro was watching her nearby. 

None of the factors above obligatorily determines the form of 
[+past, -complete] EV. It will be interesting to find out more about 
the factors that trigger a specific form. 


This work deals with a small portion of the relationship between 
embedded verb forms and tneir aspects. A toki -clause is a special case 
of a relative clause, and there is a closed set of other head nouns 
that do not leave a gap in the relative clause as their trace, such as 
tokoro 'place/time', baai 'case', koto 'thing (nominalizer) ' , no 'thing 
(nominalizer) ' , nioi 'smell', oto 'sound' and koe 'voice'. Moreover, 
subordinate structures that are marked by nara 'if, tara 'if, ba 
'if, to 'if, nagara 'while' or the te preverval-form of a verb seem 
to indicate aspect rather than tense. The diagram given in the 
previous section lacks the following case even though it should be 
possible semantically : 


EV [+st.ative, -past, -complete] S M > 


Tnis case cannot be expressed by means of a toki -clause. 

(31)*Tugino hi tenki ga waru-J^ toki EV<-MV 

next day weather bad-is 
When the weather is going to be bad on next day, 

watasi wa hana ni mizu o yar-na-i. 
I flower water give-not 
I do not water plants. 

(32)*Tugino hi tenki ga waru-katta toki EV<-M\/ 

When the weather is going to be bad on next day, 

watasi wa hana ni mizu o yar-na-i. 

I do not water plants. 


(33) Tugino hi tenki ga waru-i^ nara watasi wa hana ni mizu o yar-na-i. 

If the weather is going to be bad on next day, I do not water plants. 

Some of the subordinate-clause markers morphologically require a 
specific verb form, while some can take either the ru-form or ta-form. 
Further study is needed t extend or modify the rules for the 
toki -clause in order to capture the more general nature of aspect in 


COMRIE, Bernard. 1976. Aspect - An Introduction to the Study of Verbal 

Aspect and Related Problems . Cambridge: Cambridge University 

JOSEPHS, L. 1971. "Selected Problems in the Analysis of Embedded 

Sentences in Japanese." Unpublished Ph.D. dissertation. Harvard 

University, Cambridge. 
KUNO, Susumu. 1973. The Structure of the Japanese Language . Cambridge: 

The MIT Press. 
SIBATANI, Masayoshi, ed . 1976. Syntax and Semantics - Japanese 

Generative Grammar . New York: Academic Press. 
SOGA, Matsuo. 1983. Tense and Aspect in Modern Colloquial Japanese . 

Vancouver: UBC Press. 

Studies in the Linguist i c Science s 
Volum<? 15, Number 1, Soring 1985 


Jon Ortiz de Urblna 

The purpose of this article is to provide an analysis of 
some partitive constructions ( -rik constructions in Basque, 
genitive of negation in Russian), as an alternative to Pesetsky's 
(1982) ECP account. After reviewing some of the characteristics 
of these constructions and presenting Pesetsky's analysis, I will 
point out some of its problems. I will try to show that a 
condition on the structural relation between partitive phrases 
and the elements that define the domain where such phrases can 
appear (especially sentential negation, on which I will 
concentrate in this article, but also yes/no questions and other 
less easily characterlzable environments) accounts for the same 
distributional properties of partitive phrases as the alternative 
analysis does, while eliminating the problems related with it.l 

The analysis of partitive phrases in a language like Basque, with 
morphological ergatlvlty, is Important in that it bears directly on the 
syntactic ergatlvlty or accusativity of the language. Nominal phrases 
marked with the partitive case ending -rik may appear in the same contexts 
as phrases marked by the absolutlve ending: subjects of Intransitive verbs 
conjugated with the auxiliary izan 'to be' (11), and objects of transitive 
verbs conjugated with the auxiliary ukan 'to have' (ill). As shown in 
(1111), partitive phrases cannot appear as subjects of transitive clauses, 
where ergatlve marking is obligatory: 

(1) 1. Gizon-ik ez da etorri 
man-part neg has come 
'No man has come' 

11. Ez nuen gizon-ik ikusi herri-ko kale-etan 
neg ukan man-part see town-of street-in 
'I did not see any man in the streets of the village' 

lii.*Gizon-ik ez zuen saio-a ikusi 

man-part neg ukan perf ormance-Abs see 
'No man saw the performance' 

Since the asymmetry in distribution is not dependent on the subject/ object 
asymmetry, but rather on the absolutlve/ergative distinction, one might 
claim that partitive phrases provide some evidence that Basque shows 
syntactic constructions based on ergatlve relations. On the other hand, an 
alternative explanation that maintains the non- ergatlve status of Basque 
syntax Is possible. Levin (1983) has shown that the class of verbs 
selecting the izan 'to be' auxiliary is only a subset of the class of 
semantlcally intransitive monadic predicates, in particular, that only 
unaccusatlve verbs (in Perlmutter's (1976) terminology) belong to that 
class. It can be added as confirming evidence that unergatlve predicates 
are in fact conjugated with the transitive auxiliary ukan 'to have', their 
arguments appearing in the ergatlve case. Thus, verbs like argltu 'to 
shine', erausi 'to murmur', or kurritu 'to run' take only ergatlve 
subjects. Another sizable class of unergatlve predicates are expressed by 
[NP EGIN] phrases, treating them as particular cases of the two place 


predicate egin 'to do': Ian egln 'to work, to do work', hltz egln 'to talk, 
to do word', negar egin 'to cry, to do cry', etc. Then, following Burzio's 
(1981) GB rendering of Perlmutter's unaccusative hypothesis, izan 'to be' 
selecting unaccusative predicates would take D-object arguments while ukan 
'to have' predicates would take a D-subject only if unergative and both 
subject and object if transitive. In this analysis the asymmetry in (1) is 
a subject/object asymmetry: partitive phrases are possible in D-objects of 
unaccusative verbs (li), D-objects of transitive verbs (lii), but not 
subjects of transitives (liii). 

The partitive construction in Basque corresponds rather closely to 
some quantif icational constructions in other clearly accusative languages, 
like French pas de partitives (Kayne 1981) and Russian genitive of negation 
(Pesetsky 1.982). As discussed by Pesetsky, in clauses with sentential 
(pre-verbal) negation Russian allows genitive nominals corresponding to 
objects of transitive verbs (2i) and subjects of intransitive verbs (2ii), 
but never subjects of transitive verbs (2iii), as shown in the following 
sentences : 

(2) i. Ja ne polucal pis'ma 

I neg received letter-acc. pi. 
Ja ne polucal pisem 

letter-gen. pi. 
'I did not receive letter/any letters' 

ii. Nikakie dokladciki ne pojovilis' 

no speaker-nom. pi neg showed up-pl. 
Nikakix dokladcikov ne pojavilos' neg 
'None of the speakers/No speaker showed up' 

iii. Studenty ne smotrjat televizor 
nom. pi. watch-pl. 
*Studentov ne smotrit televizor -sg. 
'The students/*Not one student watched T.V.' 

Not all intransitive predicate subjects can appear in the genitive, 
however. As Pesetsky shows, a restriction applies to the effect that no 
genitive can correspond to the agentive subject of an intransitive verb. 
Sentence (ii) above, whose subject bears the theme theta-role, contrasts 
with (3), where the predicate assigns the agent theta-role to the subject, 
ruling out the genitive construction as ungrammatical: one place predicates 
with agent arguments fall under the unergative, rather than the unaccusat- 
ive class, and therefore agent intransitive arguments are ungrammatical 
when they appear in this partitive construction: 

(3) i. Takie sobaki ne kusajutsja 

such neg bite-pl. 

ii.*Takix sobax ne kusaetsja neg bite-sg. 
'Such dogs/*No such dogs bite' 

If theme theta-role predicates are analyzed as unaccusative verbs with a 


single D-structure object argument, an explanation for the construction can 
be supplied based on the descriptive statement that such genitives may 
appear only in D-structure object position. S-subjects corresponding to 
D-structure subjects will be ruled out. The same generalization would be 
available for Basque if verbs selecting izan auxiliary are syntactically 
unaccusative, that is, if they only theta mark, an argument in D-object 
position. In this case, partitive nominals may appear in D-structure object 

In addition to sentential negation, this type of partitive may also 
appear in other contexts like the yes/no interrogative clause below, which 
show the same distribution as their counterparts in negative contexts in 

(4) i. Arrain-ik ikusi duzu ibai honetan? 

fish-part see aux-ukan river this-in 
'Have you seen any fish in this river?' 

ii. Gizon-ik atera al da etxe horretatik? 
man-part go out Q aux-izan house that-from 
'Has any man come out from that house?' 

ill .*Pertsona-rik ulertu du nere azalpena? 

person-part understand aux-ukan my explanation 
'Has any person understood my explanation?' 

Morphologically the partitive case has one single indefinite form, as 
opposed to the other cases, which show singular, plural and indefinite 
forms. Since there is no absolutive noun in these sentences, the auxiliary 
will take the unmarked third person singular form of the absolutive marker 
as it takes the third person singular form in the Russian examples shown 

Semantically sentences with partitive -rik differ from their counter- 
parts with absolutive nominals in that the former, but not the latter, are 
given an indefinite quantification interpretation, both in negative and 
interrogative contexts. Similar to any phrases in English, the type of 
quantifier involved seems to vary between universal and existential the 
latter being common in negative and interrogative contexts. Thus, the 
interpretation of sentences (11, ii) above, with the negative quantifier, 
would be something like (5) and (6): 

(5) -' ^ X, X man, 1 saw x ( = 111) 

(6) -• 3 X, X man, x come ( = li) 

Negation has always scope over the existential quantifier in these sen- 
tences. Similarly, in interrogative contexts like those in sentences (4), 
the existential quantifier seems to be required in their interpretation. 
Corresponding to (4i,ii) we have the following: 

(7) ? 3 '^i 5^ fish, you have seen x in this river (=41) 

(8) ? 3 X, x man, x has come out from that house (=411) 


However, a universal reading occurs in other contexts where partitives may 
be used, including "affective predicate" complements and superlative 
phrases : 

(9) Ijito-rik hor bizi-tzea zoroa iruditzen zait 
gipsy-part there llve-NOM crazy seem aux 
'It seems to me crazy for (any) gipsy to live there' 

(10) Hau-xe duk. estllo-rik aurreratu-en-a 

this-emp have style-part advanced-sup-Abs 
'This is the most advanced style' 

Their interpretation is shown in the following representations: 

(11) V X, X gipsy, X to live there... 

(12) V X, X style, more (this advanced than x) 

As discussed above, only patient/theme arguments will occur in this 
construction with intransitive verbs, since only monadic verbs taking 
patient arguments select the izan auxiliary and hence surface as absolutive 
subjects, to which partitive nomlnals can correspond. 

Pesetsky's analysis tries to provide an explanation based on the ECP 
Principle. Noting that certain Russian quantifiers govern genitive case, he 
analyzes the genitive phrase as including an empty quantifier which is the 
head of the phrase, following the analysis of Kayne (1981). After applica- 
tion of Quantifier Raising to the quantifier phrase in object position, we 
get an LF representation like (13):2 

(13) [ QP [NP INFL [ V [e] ]]] 
S' i S i 

where the trace -A-bound by the quantifer adjoined to S is properly 
governed by V. However, movement out of the subject position is also 
allowed by ECP. In (14) 

(14) [ QP [ [e] INFL VP ]] 
S' IS i 

the subject trace is properly governed by the binder in adjunction 
position. If QP is not a proper binder here, the structure with subject 
QP's would be ruled out as desired. To achieve this effect, his analysis is 
based on the assumptions that categorial selection (c-selection) is 
Independent from theta-selectlon and that trace theory can be indepen- 
dently derived from other principles of UG without any specification. As a 
consequence of his first assumption, if the Projection Principle refers 
only to theta-role sub- categorization but not to c-selection, although at 
D-structure we will have a representation of the theta-role subcategoriza- 
tion frame of any verb, the constituents to which the theta-roles are 
assigned need not be of the category specified in the verb's categorial 
selection information. It is only at LF that these complements of par- 
ticular categories have to be present (and by the Projection Principle, the 
theta-roles assigned to them). By the second assumption, it follows that 
gaps are freely indexed with any constituent, whether sharing its category 


feature or not. Independent principles will rule out most cases produced by 
this free gap-Indexing mechanism. Assuming then that nothing prevents an 
empty category x from being coindexed with a c-commanding phrase of 
category y, we can therefore co-index a QP with a gap [e] of category NP 
rather than QP . In this case, if we assume that proper binding is defined 
as in (15), Incorporating condition (ill), 

(15) a properly binds b iff 

1. a and b are coindexed 
11. a c-commands b 
ill. a is a possible antecedent of b 

where for a^ to be a possible antecedent of _b they must share number, gender 
and categorlal features , then QP would not be a proper binder of the trace, 
since, as they have distinct categorlal features, QP is not a possible 
antecedent of [np e]. Then QP in -A-positlon after adjunction does not 
properly bind the trace, the latter violating the ECP Principle. At LF, 
c-selection of the verb is met since the traces are NP's, although at D- 
structure, before QR has applied, c-selectlon is not (and need not be) met 
by the QP . 

The same analysis involving an empty quantifier Is motivated for 
Basque. Although the use of partitive -rik. with overt quantifiers is 
diminishing, it is still possible, as shown in phrases like 

(16) i. Eskerr-ik asko 

thank-part many 

'Thank you / Many thanks' 

11. Hizkuntza-rik gehlen-ak 
language-part most-Abs pi 
'Most languages' 

The presence of an empty quantifier requiring partitive Case of the 
quantified element can therefore be justified for Basque as for other 
languages (Franch, Russian) where the analysis has been proposed. 3 After 
Quantifier Raising applies, the quantifier phrase adjoined to the clausal 
node is not a possible antecedent and the trace is not properly governed in 
subject position, while object positions of both polyadic ukan verbs and 
monadic unaccusatlve verbs will be always properly governed by the verb and 
QR will apply without any ECP violation being involved. 

There are some problems with the analysis, however. If partitive 
phrases are like any other quantifier phrase, we should expect to find the 
same types of scope ambiguities in clauses with more than one quantifier. 
Thus, in the following sentence, either quantifer may be interpreted as 
having wide scope over the other, allowing the two LF structures after QR: 

(17) Alkimista guztl-ek hiru sustantzla erabiltzen zuten 
alchemist all-E pi three substance use aux 
'All of the alchemists used three substances' 


i. all X, X alchemist, three y, y substance, x used y 
ii. three x, x substance, all y, y alchemist, y used x 

On the other hand, there Is one single reading for a sentence with two 
quantifiers, if one of them is the partitive phrase. For sentence (18) the 
only interpretation is as shown, with wide scope of the universal 
quantifier over the partitive: 

(18) Alkimista guzti-ek lortu al zuten urre-rik? 

all-E pi get Q aux gold-part 
'Did all of the alchemists get any gold? 

?, for all x, X alchemist, x got gold 

The same situation obtains with the negative operator, both in Basque and 
in Russian. For a sentence with sentential negation and a quantifier, like 
(19), we have the intepretations shown: 

(19) a. Ez ziren lau ikasle klase-ra etorri 

neg aux four student class-to come 
'Four students did not come to class' 

b. i. not, x=four students, x come to class 
ii. x=four students, not, x come to class 

In (i) negation has wide scope over the quantifier, and the interpretation 
is that it was not the case that four students came to class (ten came 
instead); in (ii), with the quantifier having scope over negation, the 
interpretation would be that 'for four students, it is not the case that 
they came to class (they went to the cinema instead)'. On the other hand, 
the only interpretation of a sentence with both negation and partitive, 
like the ones in (li,ii) is the one shown in (4) and (5), with wide scope 
of negation over the partitive. In addition to this problem, this analysis 
seems to make the wrong predictions with respect to the variable left 
behind by the movement of the partitive phrase at LF. If QR actually 
applies to these phrases creating an operator- variable chain at LF, the 
variable should display the same type of properties of other variables at 
that level. In particular, it should obey general principles like the 
Bijection Principle, which has among its consequences that operators cannot 
bind more than one variable. It is then predicted that application of QR to 
partitive phrases will leave behind a trace with the same coindexing 
properties as traces in cross- over structures. Thus, in (20i), the phrase 
gizon gazterik 'any young man/young men' should not be interpretable as 
coref erential with the pronoun berak 'he', since at LF the operator will 
bind both the trace and the pronoun, violating the Bijection Principle. But 
this principle holds for other Basque operator-variable relations, as shown 
in (20ii), where we have the quantifier phrase gizon asko 'many men': 

(20) i. Berak . maite duen lurra uzteak ez du gizon gazterik^ izutzen 
he . love aux country leave neg aux man youngj^ frighten 
'To leave the country that he loves does not frighten any 
young man' 


ii,*Berak. maite duen lurra uzteak ez du gizon asko . izutzen 

'To leave the country that they love does not frighten many men' 

In fact, the sentence is grammatical in the intended interpretation, 
indicating that the analysis is not adequate. 4 

The alternative analysis 1 would like to propose attempts to relate 
the two factors that seem to play a role in determining the distribu- 
tional properties of partitive constructions: on the one hand their 
limitation to D-object environments, whether as transitive object NP's or 
as unaccusative subjects, and on the other hand, their limitation to 
certain negative and interrogative domains. The intuitive idea is that 
partitive phrases have to appear within the domain of certain elements to 
be interpretable as such. I will define 'domain' here as a type of 
asymmetric c-command relationship. Thus, partitive phrases will have to 
appear 'under' sanctioning elements. The relation between the negative 
particle and the partitive phrase, will have to be one of superiority, 
where I follow Chomsky's (1973) definition according to which A is superior 
to B iff A c-commands B and B does not c-command A. Then, the negative 
element will have to be superior to the partitive phrase, following 
standard assumptions in placing NEG in INFL, in an S-structure like (21) 

(21) [ Q. INFL [VP V Q. ] ] 

the subject quantifier is not within the domain of NEG in INFL and hence it 
can correspond to a partitive phrase. Then the basic asymmetry observed 
follows directly at this level: object partitives, but not subject 
partitives will be acceptable. 

While Pesetsky's formulation crucially depends on the application of 
Move-alpha to generate operator bound variables on which the ECP account is 
based, the alternative presented here fares equally well with the movement 
and the non-movement analysis. However, if, as suggested by the evidence 
provided above, partitive phrases belong to a class of name-like quanti- 
fiers that according to Hornstein (198A) do not generate operator-variable 
structures at LF, partitive phrases will remain in their D-structure 
positions and only those within the domain of a negative element will be 
Interpretable. The ECP account is not then applicable. 

The relation between partitive and its licensing element will have to 
be checked at S-structure in this analysis. The reason is that if NEG is 
treated as an operator and moved at LF to account for scope rela- tions, 
any subject partitive would be positively sanctioned, since an operator 
adjoined to S is superior to the subject. 5 

Notice that the fact that negation has always scope over the quanti- 
fier in these constructions, which is left unexplained in the ECP analysis, 
follows as a particular case of the general relation between A and B. If 
partitives are not moved and scope relations are expressed in terms of 
c-command, as generally assumed, the negative operator will always c-com- 
mand the partitive phrase in object position. If, on the other hand, a 
movement analysis of partitives is motivated, QR could apply in any order 
to either of the elements, generating both (22) and (23): 


(22) S' (23) S- 


^ \ ^ \ 

PART . . . NEG 

Only (22) would not violate the condition on asymmetric c-coramand. Hence, 
the wide scope relation of the negative element follows directly from the 
analysis. The effect is that the asymmetry is always preserved, a not 
uncommon situation. Thus Huang (1982) motivates an analysis of scope 
interpretation where if one quantification phrase or quantifier c-commands 
the other at S-structure, it must also c-command it at LF. The cases 
discussed above would be those where the c-command relation is not 
bl-unique, and the unidirect ionality has to be preserved. 

The present analysis relies on fewer assumptions than Pesetsky's, 
while providing an explanation that is based on the relation between the 
partitive and the element that creates the semantic context for its 
appearance, a move that seems plausible. Even if all the above mentioned 
assumptions for the ECP analysis prove warranted, and ignoring the problems 
mentioned earlier, it cannot be claimed that on sinplicity grounds the ECP 
analysis involves less cost in that it follows from independent principles, 
since reference to semantic contexts 'licencing' partitive constructions 
must be made in the ECP analysis as well. 

Returning to the overall significance of these data to the analysis of 
Basque, it can then be concluded that morphological ergativity does not 
carry over to the syntax of Basque, since some apparently ergative based 
constructions can be shown to be explained in accusative terms, in ways 
parallel to other morphologically non-ergative languages. 


1 See deRijk (1972) for an exhaustive description of contexts where the 
partitive can be used in Basque. 

2 Since the head of this phrase is an empty quantifier and the case filter 
applies to phonologically realized elements only, the quantifier phrase 
does not need to be moved to subject position to receive case, as other 
elements must in Burzio's ergative verb analysis. 

3 In fact, deRijk (1972) had already proposed an analysis based on empty 
quantifiers in Basque 

4 Hornstein (1984) considers quantifiers like any in English as belonging 
to a name-like class that does not create operator-variable structures at 

5 This is not an implausible assumption. Other relationships like Kayne's 
connectedness relation are also assumed to be checked at that level. 


BURZIO, L. (1981). Intransitive verbs and Italian auxiliaries , 

Doctoral dissertation, MIT, Cambridge, MA 
CHOMSKY, N. (1973). "Conditions on Transformations", in S. 

Anderson and P. Kiparksy (eds.) 1973, A Festschrif t for 

Morris Halle , Rinehart and Winston, New York 
HORNSTEIN, N. (1984). Logic as Grammar , MIT Press, Cambridge, MA 
HUANG, J.T. (1982) Logical relations in Chinese and in the 

theory of grammar, Doctoral dissertation, MIT, Cambridge, MA 
KAYNE, R. TT981). "ECP extensions". Linguistic Inquiry 13, 

LEVIN, B. (1983). On the nature of ergativity . Doctoral 

dissertation, MIT, Cambridge, MA 
PERLMUTTER, D. (1978). "Impersonal passives and the Unaccusative 

Hypothesis", in J.J. Jaeger and A.C. Woodbury (eds.). 

Proceedings of the Berkeley Linguistic Society , 4 , 

Berkeley, CA 
PESETSKY, D. (1982). Paths and Categories , Doctoral dissertation, 

MIT, Cambridge, MA 
deRIJK, R. (1972). "Partitive assignment in Basque", in 

Anuarlo del Seminarlo de Filologla Vasca 'Julio de 

Urquijo .VI, 130-173 

studies In th< LingulBtlc Sciences 
Volume 15, Nuaber 1, Spring 1985 

Hyang-Sook Sohn 

The irregular verbs of Korean under discussion exhibit 
alternations between null and obstruents (/s/-irregular 
verbs), £ and t^ (/t/-irregular verbs), u, w, and £ 
(/p/-irregular verbs), £ and null (/-fZ-irregular verbs), 1^ 
and null (/l/-irregular verbs), and 1_+ and 1_1 
(/l+/-irregular verbs). The alternation of the stem-final 
segment before vowel-initial suffixes has interested Korean 
phonologists , but there are still shortcomings in the 
previous accounts. In this paper, I argue that the vertical 
encoding of multi-level representation and the horizontal 
encoding of cyclic expansion accounts for the deeper 
phonological regularity of the irregular verbs, which would 
otherwise not be captured. In nonlinear analysis, rules 
apply to both regular and irregular verbs only if the 
structural description is met, and no diacritic is needed to 
differentiate these two types of verbs. 

1. Introduction 

Korean has two classes of verbs: those whose stem-final segments show 
alternations before some suffixes ("irregular verbs"), and those whose 
stem-final segments do not show any alternations throughout the verbal 
paradigm ("regular verbs"). The alternation of the stem-final segment 
before vowel-initial suffixes in the irregular verbs has been of interest 
to Korean phonological studies and linear analyses have been proposed to 
account for it (Kim 1971, Kim-Renaud 1973, Lee 1976). 

The present analysis, however, provides a simpler and more insightful 
account of these alternations by resorting to the recent developments of 
the phonological theory: first, the nonlinear phonology allows phonological 
rules to have a variety of reference by separating the concentrated 
information in a single linear sequence into several levels and postulating 
the skeleton and the melody as autonomous representations (Clements and 
Keyser 1983, Harris 1983, and Levin 198'*). Secondly, the lexical model of 
the interaction between phonology and morphology (Mohanan 1981, Kiparsky 
1982, 1983, and Pulleyblank 1983) allows phonology to have a more direct 
access to morphology. A third theoretical assumption is the theory of 
underspecification (Kiparsky 1982, 1984, Archangeli 198**). Assuming 
underlyingly unspecified segments allows rules to be formulated in a 
simpler way. These theoretical frameworks provide the phonological 
component with an enriched source of phonological differentiation. 

In the next section, an outline of the theoretical framework assumed 
In the present analysis Is sketched. The following section provides an 
overview of regular verbs to show that they show alternations which are not 
restricted only to the verbal phonology of Korean. In the final section, 
we discuss the alternations peculiar to the Irregular verbs and argue that 
they are best treated in terms of nonlinear phonological representations In 


which irregular verbs are different from regular verbs. 

2. A Sketch of the Theoretical Framework 

In this section, I will briefly outline the major theoretical 
frameworks on which the analysis in the following sections are based. The 
essential assumptions concerning the syllable structure of Korean, 
underspecif ication of features, and the model of lexical phonology are 
discussed . 

2.1. Syllable Structure 

The analysis in this paper makes crucial reference to the skeletal 
tier and the syllable structure built on it. Following Levin (198'4), the 
skeletal tier consists of a sequence of empty x slots which represent 
quasi-timing units. Each feature matrix in the segmental tier is 
categorially classified with [+N] or [-N]. In the underlying 
representation, however, only [+N] is specified. [+N] segments are 
obligatorily syllable nuclei. Even though a segment is unspecified with 
respect to the categorial component, it can be assigned the value [+N] if 
the segment is under the nucleus of a syllable. 

(1) Nucleus [+N] Assignment 

[ ] > [+N] / 

Nucleus [+N] Assignment (1) applies to a newly created nucleus x slot which 
is derived by rule. Otherwise, [+N] is specified underlyingly. Otherwise, 
a feature matrix is assigned a default value [-N]. 

(2) Default [-N] Assignment 
[ ] > [_N] 

The two rules (1) and (2) are intrinsically ordered by the Elsewhere 
Condition. Since the environment of rule (1) is more specific than that of 
(2) and the outputs are distinct, rule (1) applies prior to rule (2). 

In Korean all the vowels except ^ and u are assigned [+N] 
underlyingly. All consonants and i^ and u are underlyingly unspecified with 
respect to the categorial feature. Consonants are assigned [-N] by Default 
rule (2). The two vowels i^ and li which alternate with glides ^ and w 
respectively are assigned [+N] by Nucleus [+N] Assignment (1) when they are 
dominated by the nucleus. Otherwise, they are assigned [-N] by rule (2), as 
consonants are. 

The syllable structure built on the skeletal tier contains information 
on the categorial specification of each slot. The syllable structure of 
Korean can be roughly represented as (3). 




Universal grammar specifies the position of the syllable head, N. The 
nucleus N node percolates itself up to N" , which is the highest constituent 
for a syllable. Thus, N" cannot exist without the syllable head, N. The 
stray x slots are incorporated into the syllable constituent node to build 
a syllable. 

(4) Syllable Building Rules 

a. N" Branching 

Attach to the left of N" the x' immediately preceding x (x' 
refers to the unsyllabified x ), 

b. N' Branching 

Attach to the right of N' the x' immediately following x. 

The application of the rules (la) and (4b) are shown as (5a) and (5b) 

(5) a. 





X x' > X X 

I I I 

[+N] [>N]t-N] 

Rules CJa) and (Ub) apply in that order. Core syllabification of Korean 
represented by rule (4) allows at most binary branching for N, N', and N" 

2.2. Underspecification 

It is assumed in this paper that the syllable structure and vowel 
melody planes involve underspecification in their underlying representation 


(Archangel! 198^). Universal grammar specifies nothing but the nucleus 
position. Other parts of the syllable structure are built by either 
language particular or universal parameters. Thus, the three kinds of 
nuclei in Korean, short vowel, long vowel and diphthong are represented 
with the following underlying structure: 

(6) a. N b. N c. N 

I A A 


i V II 

[ ] [ ] [ ][ ] 

(6a) represents a nucleus N occupying a single slot i.e. a short vowel. In 
(6b), a single feature matrix occupies two skeletal slots under N, and 
represents a long vowel. In (6c), however, two slots under a branching N 
are linked to different features i.e. a diphthong. The contrast between 
(6b) and (6c) illustrates how a long vowel is differentiated underlyingly 
from the diphthong. Syllable Building rules (4) and the underlyingly 
specified nucleus structures in (6) predict that a syllable in Korean 
occupies at most four x slots. 

The assumption of the underspecification of the vowel melody plane is 
crucial to the simplification of the grammar. The fact that * w-» and * y+ 
are not possible diphthongs and that the featureless nucleus surfaces as a 
high back unround vowel [+] suggests that the vowel [+] is least marked in 
Korean. If we follow Archangeli (^98'^) in assuming that the redundancy 
rules which fill in the feature matrices interact with other phonological 
rules, these two observations are accounted for in a much simpler way, as 
briefly mentioned below. 

The absence of diphthongs including + as a half of their structure is 
simply stated as deleting the x slot which has no feature matrix, as will 
be shown with respect to (^Deletion (25). If the vowel matrix is fully 
specified in the underlying representation, the gaps in diphthongs can be 
accounted for only by stipulation. We return to this in the discussion of 
the £-irregular verbs in U.3. Considering the interaction of the vowel ^ 
represented as an empty nucleus x slot with other vowels, underspecifying 
the features and later filling in the unspecified features by redundancy 
rule is less costly than specifying the inserted x slot with concrete 
features [+high, +back, -round] from the underlying entry. We return to 
this in the discussion of the vowel-final regular verbs in 3-'^' Though a 
full-fledged account of the underspecification of Korean vowels is not 
relevant to this paper, it is necessary to see where the features for the 
vowel + are assigned. As Baker (1982) argues, the feature [-back] is 
underlyingly specified in Korean. A redundancy rule fills in the default 
value [+back] to the unspecified slot after certain phonological rules. 
Finally, default values [+high] and [-round] are inserted by Complement 
Rules (Archangeli 1984), spelling it out as [+]. The underspecification of 
the syllable structure and of the feature matrix for a vowel simplifies the 
grammar and provides an explanatory account of the phonological 
alternations. This is illustrated with respect to vowel deletion, 
monophthongization, and glide formation rules in lexical phonology. 

As already mentioned, the account of consonant alternations in this 


paper makes crucial use of the lexical stratum of verbal morphology. We 
now turn to an outline of the lexical model of Korean phonology. 

2.3. Lexical Phonology 

Korean has a rich morphology of verbal suffixes. The verbal form adds 
its suffixes from the verbal stem to the right in an endocentric fashion. 
Illustrative among the verbal suffixes are morphemes of honorifics, 
formality, and aspectuals. For example, the verbal form koQpuha-si-pni-ta 
expands as in (7), where C stands for "cycle n". 

(7) [[[[verbal stem] honorific] formal]_ indicative ending] „ 

[[[[ kogpuha ] si] "^ £ni] ^^ ta] ^^ 

Lexical phonology allows the application of phonological rules to follow 
each morphological operation. That is, each addition of morphological 
structure is followed by application of phonological rules. Thus, the 
morphology and phonology processes are interwoven. 

In order to properly account for the alternation of the final 
consonant of the irregular verbal stem, it is necessary to look into the 
morphological stratum in addition to the cyclic expansion shown in (7). 
Following the model of lexical phonology (Mohanan 1981), I assume the 
lexicon of Korean verbal morphology consists of ordered lexical strata, as 
shown in (8). The cycles C„ and C in (7) belong to a single stratum, and 
the aspectuals constitute a separate stratum. 

(8) Stratum I: Nominalizer m 

Effective marker ni^ 
Honorific marker si^ 
Propositive marker psl 
Formal marker spni 
Stratum II: Indicative ta 

Connective ko 

Statlve a 

There is no loop between stratum I and stratum II in Korean. That is, 
the stratum II morpheme cannot feed back the stratum I morpheme. 

The expansion of each cycle to add a morpheme is followed by the 
application of the phonological rules available in the corresponding 
stratum. The lexical rules that apply in the domain of a stratum meet the 
Strict Cycle Condition (Mascaro 1976) because they should be prevented from 
applying in a non-derived environment. The lexical rules apply across 
brackets or if the environment is created by application of a phonological 
rule. Contrary to the lexical rules which are sensitive to word internal 
morphological information, postlexical rules do not refer to the internal 
constituent structure since all brackets are erased at the end of each 
stratum, although not at the end of each cycle (Mohanan 1981). 

The model of lexical phonology, along with underspecification, 
contributes to simplifying rule formulation. If the redundancy rule which 
assigns a feature matrix for the least marked vowel £ in Korean is delayed 
as a postlexical rule, both lexical and postlexical rules which are 


sensitive to the vowel *_ turn out to refer to the skeletal tier, regardless 
of the features that are linked to the x slot. The +-<Jeletion rule, for 
instance, can be simply formulated to delete a featureless x slot. 

With the theoretical assumptions and the lexical model of Korean 
verbal morphology sketched so far, we now turn to the analysis of the 
verbs. Before discussing the irregular verbs, I provide an overview of the 
regular verbs in Korean in the following section. 

In the analysis of the verbs, morphemes are assumed to have the 

following underlying representations. The x slot with a vertical line 

above it refers to a nucleus specified in the underlying representation, 
whereas the simple x slot a consonantal position. 

(9) a. Norn. b. Effect, c. Honor. d. Propos. e. Formal. 

XX XXX xxx xxxx xxxxx 


m ni si psi spni 

e. Ind. f. Conn. g. Stat. 

I I I 

XX XX x 


t a k o a 

The underlying representations above account for the alternation between 
the presence and absence of the vowel + which is represented by the 
unspecified nucleus x slot. We will return to this in 2-'^' 

3. Regular Verbs in Korean 

Study of the regular verbs reveals how regular the behavior of the 
irregular verbs are. In what follows we look into regular verbs which are 
relevant to the discussion of the irregular verbs. Among the verbs with a 
consonant-final stem, /s/~, /t/-, and /p/-regular verbs are of particular 
interest since they contrast with their irregular counterparts. Verbs with 
vowel-final stems are also interesting due to their interaction with the 
following suffix initial vowel. 

3.1. /s/-Regular Verb 

/s/-regular verbs refer to a class of verbs whose stem ends with the 
consonant £. 

(10) Nom. Effect. Honor. Propos. Formal 

a. p*s-+m pd3-+ni pas-+si pas— J-psi pat-s'+pni 

b. s'is-+m s'is-+ni s'is-+si s'is-+psi s'it-s'+pni 

c. us-+m us-+ni us-+si us-+psi u:t-s'+pni 

Ind, Conn. Stat. Gloss 

a. pat-t'a p^k-k'o pas-a "to wear off 

b. s'it-t'a s'ik-k'o s'is-a "to wash' 

c. u:t-t'a u:k-k'o us-a "to smile' 


In the verbal paradigms in which the suffix begins with the nucleus x slot 
i.e. with vowel, the stem-final consonant £ does not show any alternation. 
In the formal, indicative, and connective forms, however, the stem-final 
consonant £ alternates with either t or k. In the former case where the 
stem-final £ does not show any alternation, all that is left for phonology 
to do is to fill in the feature matrix for the unspecified nucleus x slot 
in the suffix. Thus, the honorific form p?s— i-si is derived by inserting the 
vowel [+]. 


N" N" "^ 

N" N" N" 
/N' /N' /^' 



pas [+]s i 

The feature matrix of an empty x slot is determined by the redundancy rule 
if the X slot is directly dominated by N node in the syllable structure. 
Since the empty x slot in (11) is the nucleus of the syllable, it is 
underlyingly specified with the categorial feature [+N]. Once the x slot 
is specified with respect to the categorial feature [+N], it is subject to 
the redundancy rule for a [+N] segment. Redundancy rules fill in the 
feature matrix for the unspecified nucleus x slot. 

(12) [+N] Redundancy Rules 

a. [ ] — > [fhigh] 

b. [ ] ~> [+back] 

c. [ ] — > [-round] 

d. [ ] — > [-low] 

[■••N] Redundancy Rules in (12) spell out the empty nucleus x slot with the 
least marked vowel [ + ], thus deriving pas— t-si in (11). 

In the latter case where the stem-final £ alternates with either t or 
k, the phenomena involved in the alternations are accounted for by such 
phonological rules as Tensing, Neutralization, and Assimilation. 

Tensing of the suffix-initial obstruent is explained by a very general 
rule in Korean. 

(13) Tensing 

X — > [ttense] / x 

I I ~ 

[-son] [-son] 

The second obstruent in obstruent clusters is always tensed by the 
preceding obstruent. For example, /kakca/ ^individual' is realized as 
kakc'a , and /c»pta»' ^entertainment' as opt 'afc Since Tensing (13) applies 
in a nonderived as well as derived environment it is a postlexical rule. 
The alternation of the suffix initial consonants s, t, k with s', t', k' in 


formal, indicative, and connective forms respectively in (10) do not form a 
peculiar part of the Korean phonology. 

Neutralization is another postlexical rule by which the obstruent in 
the coda position is neutralized into the homorganic stop. 

(14) Neutralization 

-son 1 r -son "| 
+contJ > L-co"tj 

The formal form p9t-s'-»pni is derived as follows: 


r N" 

N'n Red. (12) 


I I I 

LU) a s 



I I I 

pat s' [+]p n i 

The connective form pak-k'o involves an assimilation of place of 

(16) Assimilation 


r+cori f-cor"\ 
L+antJ b^antJ 

The assimilation of the alveolar obstruents into labial or velar is 
another common postlexical rule since it applies in nonderived environments 
as well. For example, /satpulli/ ^hastily' is realized as sapp'uli , and 
/catkarak/ ^chopsticks' as cakk'arak . 

The derivation of pak-k'o is as follows: 

p ^ t 

The derivation of the verbal forms of the /s/-irregular verb involves 
[+N] Redundancy Rules (12), Tensing (13), Neutralization (1«), and 


Assimilation (16), all of which apply in the postlexical cycle. One 
alternation left unaccounted for is that of stem vowel length in (10c), 
is discussed in 3.3. 


The reason the verbs in ( 10) are regarded as regular in spite of the 
alternations in the formal, indicative, and connective forms is that the 
alternations are general phenomena with respect to the whole phonology of 
Korean. That is, the four rules needed to account for the verbal 
alternations are not peculiar only to the verbal forms. 

3.2. /t/-Regular Verbs 

The verbs in (18) show some of /t/-regular verbs. 












mut— J-m 

mut— i-ni 

mut— +-si 















^to pull up' 





^to bury' 





'to tread' 

Formal . 

The alternation of the stem final t with Ik before the connective suffix ko 
is accounted for by Assimilation (15). The connective form k»k-k'o is 
derived as follows: 







I I I 

Uk a t 





X X 

I I 
k o 

Tens. (13) 

Assim.( 16) 

The alternations involved in /t/-irregular verbs are accounted for without 
adding any more rules to the phonological component. 

3.3. /p/-Regular Verbs 

/p/-regular verbs refer to the verbs whose stem-final consonant ends 
with £. 








kop— i-m 

kop— J-ni 


kop— i-psi 




kup— J-ni 


kup— i-psi 



s'ip— Hn 

s'ip— J-ni 



s' i:p-s'+pni 









*to be numb' 





'to bend' 


s'i:p-t'a s'i:p-k'o s'ip-a 

'to chew' 

Tensing of the suffix-initial obstruent is, as already pointed out, a 
common phenomenon. However, the alternation of the stem vowel length in 
(20c), along with the one in (lOc), remains to be explained. A closer look 
at the verbal forms in (20c) reveals that the length of the stem vowel is 
dependent upon whether the following suffix begins with vowel or consonant. 
That is, the underlying long vowel in the stem becomes short when the 
following suffix begins with vowel. 

(21) Vowel Shortening 


(x) ^ 

Vowel Shortening (21) is a lexical rule since it refers to the morpheme 

The derivations of s'ip— i-ni and s'i:p-s'^^pni are as follows: 


N" N" 

^N' /N' 

In (22b), Vowel Shortening does not apply because the suffix does not begin 
with syllable nucleus. 

Shortening the stem vowel is a general process in verbal phonology, 
conditioned by the suffix initial vowel. As we will see below, it applies 
regardless of whether the verb is regular or irregular, and regardless of 
whether the stem is consonant- or vowel-final. 

3.4. Verbs of Vowel-Final Stem 

Verbs whose stem ends with vowel show interesting phenomena with 
respect to underspecification. 














































"to lose' 





^to put on 

the head' 





"to go' 





"to stand' 





"to cut' 

The verbs in (23) are regular in the sense that the verbal stem does not 
alternate through the verbal paradigms. However, the suffix-initial 
nucleus is systematically deleted when the suffix begins with a syllable 
nucleus. The other alternations to be accounted for are deletion of the 
initial consonant £ in the formal suffix, and the absence of the stative 
suffix in (23c, d). 

In case of the vowel-final verbal stem, suffixation results in a 
sequence of syllable nuclei when the suffixal morpheme begins with a 
syllable nucleus. One of the abutting vowels in such a sequence is always 
deleted in stratum I. Let's take, for example, the effective form ci-ni in 
(23a). The morphological derivation in stratum I gives the following 
detailed representation. Universal Association Convention associates the 
feature matrices with x slots in the skeletal tier one to one from left to 



. c [-back] „ 



n [-back] 


In (2M), one of the two adjacent vowels is specified with the feature 
[back], while the other is totally unspecified. In a sequence of two 
syllable nuclei, the unspecified syllable nucleus Is deleted. 

(25) 6-: 



— > /a / 

Circled x slot with a vertical line refers to a nucleus vowel whose 
■feature matrix is empty. Empty Nucleus Slot Deletion (25) is a lexical 
rule since it refers to a bracket. Thus, the unspecified nucleus x slot In 
(24) is deleted by (^-Deletion (25), and the correct form ci-ni is derived. 

In the stative forms in (23) where the two vowels surface, however, 
neither the stem-final vowel nor the stative suffix £ is unspecified. 
Thus, the structural description of g>-Deletion (25) is not met, and both of 

the two adjacent vowels surface, as in ci-a , i-a , and pe-a . 

The appearance of vowel sequences in the stative forms in (23) 
provides evidence for underspecif ication . It shows what counts in deletion 
of a vowel in the vowel sequence is not linear order but the presence or 
absence of specified features. If all the distinctive features are 
under lyingly specified for all the vowels as SPE theory (Chomsky and Halle 
1968) does, then the vowel -i- is no different from the rest of the vowels in 
Korean, As a consequence, deletion of the vowel +, but not of any other 
vowel, in a vowel sequence is accounted for by an ad hoc stipulation to 
that effect. Following underspecif ication , however, a vowel sequence with 
the vowel + is not comparable to a sequence without +_^ The former has an 
empty x slot, while the latter has no empty x slots. Since the two types 
of nucleus sequences are different in their featural representations, it 
does not take an ad hoc stipulation for the unspecified nucleus x slot to 
be deleted. Rather, it naturally follows from the theory of 
underspecif ication , according to which +^ is asymmetrical to the rest of the 
vowels in Korean, 

Turning to the formal form ci-pni in (23a), we need a rule to delete 
suffix-initial s after a stem-final vowel. 

(26) ^-Deletion 

X > ^ / 


s (Domain: Stratum I) 

The formal form ci-pni in (23a) is derived as follows; 






c i ; 

p n 1 J 

p n 1 ^ 

c i p n i 

In the derivation in (27), application of ^-Deletion (26) feeds ®- 
Deletion (25). This is why the formal suffix loses its empty nucleus slot 
although it does not begin with empty nucleus x slot as other suffixes in 
stratum I do. 

One more point to be made about the verbal forms in (23) is the vowel 
deletion in the stative forms of (23c, d). Since the stative suffix is a, 
^Deletion (25) is not applicable. Deletion of the stative vowel is simply 
a contraction which applies when the two adjacent vowels are of the same 


(28) Contraction 

N N 

X > / X 

I 1 

To illustrate, ka as stative form in (23c) is derived as follows: 

(29) rf N "^ N" ^ 
/ I 


k a ^ 

Although paradigms of only five vowel-final verbs are illustrated in 
(23), the vowel-final verb behaves regularly whatever vowel the stem ends 
with, except i^. Discussion of the i^-final stem is given in 4.5. 

3.5. Summary 

The regular verbs in Korean need phonological rules to account for the 
interaction between the verbal stem and the suffix. Among the postlexical 
rules are [-t-N] Redundancy rules (12), Tensing (13), Neutralization (14), 
and Assimilation (16). Among the lexical rules are Vowel Shortening (21), 
(^Deletion (25), ^-Deletion (26), and Contraction (28). These lexical 
rules, except s-Deletion (26), do not need a lexical domain specified as a 
part of their rule description. All the postlexical rules listed above are 
general rules in the sense that they are required to account for general 
Korean phonological processes. The lexical rule (^Deletion (25) is also a 
general rule since it applies to nouns when they are followed by affixes 
which begin with a featureless nucleus. On the other hand, the rest of the 
lexical rules apply only to the verbs. 

With the rules necessary to account for the regular verbs, we now turn 
to discussion of the irregular verbs to show that they are no less regular 
than their regular verbs. 

The Irregular Verbs 
1 . /s/-Irregular Verb 

In figure (30) we see some verbs traditionally called /s/- irregular 
. This class of irregular verbs contrasts with the regular 


counterpart in (10). 

















i— J-psi 



k-i— +m 


k-f- 4-sl 

k+- +psi 




Conn . 






^ to make' 




^ to connect' 




^ to draw' 

Of interest first is the vowel-final to consonant-final alternation. 
Another strange alternation to be accounted for in the irregular verbal 
forms in (30) is the appearance of a sequence of vowels across the morpheme 
boundary. In discussion of regular vowel-final verbs, it has been noted 
that the verbal forms in Korean do not tolerate vowel sequence in derived 
environments if one of the vowels is /+/. Recalls-Deletion (25) here. 
None of the regular vowel-final verbs in (23) allows two adjacent vowels, 
while the /s/-irregular verbs in (30) do. 

The non-application of ©-Deletion (25) to the so-called /s/- irregular 
verbs in (30) is reduced to two possible reasons: one is that the stem of 
this class of verb is not vowel-final, and the other is that it is 
vowel-final but this class of verbs are diacritically marked so that 
^-Deletion (25) cannot apply. In accounting for the appearance of vowel 
sequences in (30), we prefer the former to the latter because a diacritic 
use of a phonological rule does not provide any explanatory adequacy. 

Now, if the appearance of vowel sequences in (30) is due to the fact 
that the stem does not end with vowel, then it must be the case that either 
the stem is consonant-final or it has some featureless consonant slot in 
the stem-final position. In case the stem is consonant-final, we can 
derive the correct form, for example, ci— »ni in (30a), by deleting the 
postulated stem-final consonant and ordering ^Deletion (25) prior to the 
stem-final consonant deletion. Since the two processes are extrinsically 
ordered in a bleeding relation, we derive the correct form with a vowel 
sequence. However, this analysis is not plausible because the features of 
the postulated stem-final consonant are totally arbitrary. At best, 
deletion of the stem-final consonant is diacritically marked since there 
are regular verbal counterparts whose stem-final consonants are not 

Adopting the alternative possibility, if the stem of the so-called 
/s/-irregular verb ends with a featureless x slot which is redundant with 
respect to the feature matrices in the segmental plane, then 
non-application of ©-Deletion (25) is exactly what is expected since the 
suffix-initial nucleus is not abutting the nucleus in the stem due to the 
intervening stem-final featureless x slot. 

Comparing the /s/-irregular verbal forms in (30) with the /s/-regular 
verbs in (10) and vowel-final regular verbs in (23), I suggest different 
underlying representations. 

(31) a. N (30) b. N (10) c. N (23) 

A I I 


11/ III II 

c i P d s c 1 

(31a), (31b), and (31c) are the underlying representations of the verbal 


stems in (30a), (10a), and (23a) respectively. What is noticeable is that 
the so-called /s/-irregular verbs have an empty x slot in the skeletal tier 
in the underlying representation. Core syllabification attaches it to the 
N' node. n' 



I V 


Thus, (31a) is differentiated from (31b) with respect to whether the final 
slot under N' is segmentally empty or not (interested readers are referred 
to the Turkish data in Clements and Keyser (1983) for a similar case), and 
it is differentiated from (31c) with respect to whether the N' node is 
branching or not. It has already been explained in 3.** that a vowel is 
deleted when abutting another vowel, and that the precise determination of 
the vowel to be deleted is predictable on the ground of under specification. 
These facts are captured in (x>.Deletion (25), which is repeated below for 

(25) ©-Deletion 

N 1 

X > / X j 


At first glance, the stratum I verbal forms in (30) (except for the 
formal forms) look like counterexamples to (X) -Deletion (25), thus turning 
out to be diacritically marked as exceptions to the rule. However, it is 
argued that the structural description of (J>-Deletion (25) is not met in 
those forms if we assume the nonlinear underlying representation like 

The effective forms ci-4-ni and ci-ni respectively in (30a) and (10a) 
are differentiated with respect to the applicability of (^Deletion (25). 
The derivations of ci— »ni and ci-ni are shown in (33a) and (33b) 

(33) a. 

N" N" '^ VS(21) b. ^r N" 
/N' Aj' Red(12) 


Del (25) 

In (33a) the empty x slot which makes the N' node branching blocks the 
application of the (^Deletion (25). That is, the empty x slot intervenes 
between the two otherwise adjacent nuclei. Unless the skeletal tier is 
separately represented from the feature matrix, the empty x slot is 
invisible because it has no feature. As a result, the (^Deletion (25) 
would apply in (33a) as it does in (33b), deriving an incorrect form for 


(33a). Within the framework of nonlinear representation, however, the 
structural description of (^Deletion (25) is simply not met in (33a) since 
the rule deletes an unspecified nucleus x slot immediately after another 
syllable nucleus. 

The empty x slot is resyllabif ied as the onset of the following 
syllable at the end of the lexical cycle. The x slot which is assigned no 
feature is not realized phonetically. It is shown shortly that the empty x 
slot in the coda position is specified with features by rule. L+N] 
Redundancy rule (12) derives ci— »ni in (33a). In (33t>) , (x>-Deletion (25) 
derives ci-ni . 

It is only by way of incorporating the skeletal tier in the underlying 
representation that the two forms in (33) are distinguished. The 
appearance of a sequence of vowels, though one of the vowels is the 
underlyingly unspecified vowel +, is not an exception to the application of 
(^Deletion (25). Since the inapplicability of gl-Deletion (25) in the 
/s/-irregular verb naturally follows from the underlying representation 
like (31a), it is incorrect to regard this class of verb as irregular. The 
least marked vowel i_ appears though preceded by another vowel, because it 
is not abutting the preceding vowel in the skeletal tier. 

Let's look at the formal, indicative, and connective forms in (30). 
We find that the empty skeletal slot in the underlying representation is 
filled only when a consonant-initial suffix follows. The empty slot in the 
stem is filled by a redundancy rule for [-N] segments. 

(.3'^) [-N] Redundancy Rule 
N' N' 


"N"' in the formulation of rule (34) refers to a consonantal position in 
rime i.e. a coda. If [-N] Redunaancy Rule (3'<) applies on any cycle, it 
fills in the empty x slot in every verbal form of the /s/-irregular verb, 
which is not correct. In order to derive the correct form, we follow 
Archangeli (1984) in assuming that the redundancy rule applies as late in 
the postlexical cycle as possible, as the [+N] Redundancy Rules (12) do. 

The connective form in (30a) is derived as follows: 

(35)r'r N" -T N'^ [-N]Red. (34) N" N" 

. k o J 

Tens. (13) 



Assim.( 16) 

X X X X 

i V 1 

c i k 

X X 

1 1 

In the derivation of the verbal forms, resyllabif ication is assumed to 


apply at the end of every cycle. In the lexical cycles, no rule is 
applicable. In the postlexical cycle, the empty x slot is attached to the 
N' node as coda of a syllable. Thus, [-N] Redundancy Rule (34) applies, 
filling in the empty x slot with [t]. The correct connective form ci:k-k'o 
is derived by Tensing (13) and Assimilation of place of articulation (16). 

The derivation of the formal form ci;t-s'-»-pni in (30a) is shown in 
(36). The lexical entry at stratum I is (36a). 

(36) a. 




X X X X 

I V 

c i 



X X X X X 

p n i j 

[-N] Red(3'») 

The empty x slot which makes the N' node branching blocks the application 
of ^-Deletion (25). As a consequence, the verbal stem is not directly 
followed by a vowel, and hence Vowel Shortening (21) does not apply. The 
empty x slot is not resyllabified as onset of the following syllable, 
allowing [-N] Redundancy Rule (34) to apply (compare with the derivation of 
vowel final verbs in (27) where there is no empty x slot). 

The derivations in (35) and (36) are no less regular than regular 
verbs, since their underlying representations are different from those of 
the regular verbs. [-N] Redundancy Rule (34) is the only one that is 
additionally motivated to account for the irregular verbs. But this may be 
independently motivated as a more general rule in Korean . 

The traditional analysis, in which s^ appears underlyingly as a 
stem-final consonant, still needs Neutralization (14) to derive t in formal 
or indicative forms. Therefore, adding one more rule in the present 
analysis does not simply make the grammar more complex. Rather, the present 
analysis shows that the /s/-irregular verbs behave differently from the 
/s/-regular verbs because the underlying representations of these two 
classes of verbs are different. As a result, the alternations in 
/sZ-irregular verbs are not unpredictable or idiosyncratic, but entirely 
rule-governed and regular. 

To recapitulate, the vowel- to consonant-final alternation on the one 
hand, and the systematic appearance of a vowel sequence with the vowel + 
preceded by stem-final vowel on the other, are best treated by assuming an 
underlying representation like (31a), namely, where the stem looks 
vowel-final in the segmental plane but actually is not vowel-final in the 
core skeleton. 

So far, assuming an empty x slot in the underlying representation of 
the /s/-irregular verb has been motivated with respect to (J^-Deletion (25) 
and ^-Deletion (26). Another piece of evidence that the stems of the 
/3/-irregular verbs have an empty x slot in the skeletal tier is provided 


with respect to the glide formation rule which applies in the postlexical 

(37) Glide Formation^° 
N N 
t .-'1 


[ +sonJ 

In a sequence of two vowels where the first vowel is _i or u_ and the second 
one is not i, the high vowel becomes a glide before another vowel. The 
application of Glide Formation (37) is optional. Thus, the stative forms 
in (23) show the applicability of Glide Formation (37). But rule (37) 
cannot apply to the stative forms of the /s/-irregular verb in (30). To 
show the contrast: 

(38) a. ci-a/ cya ^ to lose' i-d/ ya Ho carry on the head' (23) 
b. ci-a/ "cya Ho make' i-a/ "ya Ho connect' (30) 

The two forms in (38a) are the stative forms of the regular verbs in (23), 
while those in (38b) are the stative forms of the irregular verbs in (30). 
Since the stative forms in (38b) have an empty x slot intervening between 
the syllable head of the stem and that of the stative morpheme, it does not 
meet the structural description of Glide Formation (37). Unless we refer 
to the skeletal tier, this distinction cannot be made since the empty x 
slot is not defined by featural representation. 

The way in which the verbal forms in (30) interact with (g-Deletion 
(25), ^-Deletion (26), and Glide Formation (37) provides substantial 
evidence for postulating structures like (31a) as the underlying 
representation of the /s/-irregular verb. Thus, the presence or absence of 
a consonant in the stem-final position, and the presence of the two vowels 
in a row in the /s/-irregular verb are accounted for by making crucial 
reference to the skeletal tier and the underspecified features. With 
respect to the account provided in the present analysis the so-called 
/s/-irregular verbs are neither irregular, nor do they end with s in their 

4.2. /t/-Irregular Verb 

This class of verbs is called /t/-irregular because the stem-final 
consonant alternates between t and r. 








kar— i-m 



k»r— ipsi 



mur— »-m 


mur— I-si 

mur— i-psi 



t+r— i-m 















Ho walk' 







^to ask' 





Ho hear' 


Except for the connective, the alternation in (39) is reduced to that 
between t^ and r. If we take the underlying phoneme of the alternating 
segment as /r/ , the words in (39) are derived rather straightforwardly. 
The underlying phoneme of the stem-final consonant in the regular verb in 
(18) is /t/. 

(HO) Underlying Representation 

a. /t/-regular b. /t/-irregular 

N N 


*to pull up' 


X X X X 

k » r 

*to walk' 

Assuming the underlying representations like (40b), the derivation of 
the verbal forms in (39) whose suffix begins with a vowel is simply without 
any alternation. The derivation of the prepositive form kar— i-psi is as 

(11) r- 

VS (21) 

[+N] Red. (12) 

[+]p 3 i 

The stem-final consonant r_ alternates with t in formal, indicative, 
and connective forms in (39). The position of r_ in the syllable in these 
forms is different from that of the non-alternating £. In the former case, 
£ remains in the coda position, while in the latter case, it is 
resyllabified as the onset of the following syllable since the suffixes 
begin with vowels. 

In general, Korean [r] is not regarded as a phoneme, but is an 
allophone of the phoneme /I/, which is distributed to appear only in the 
onset position. This explains why stem final r_ does not alternate before 
vowel-initial suffixes but does alternate before consonant- initial 
suffixes. £ in the coda position is not viable in Korean and it is 
deleted. -^ 

(42) r-Deletion 

> ff I 

Since £-Deletion (42) refers only to the segmental plane, the x slot for r_ 
remains empty. Take, for instance, the connective form ka;k-k'o in (39a). 



H" ^ 

N" 1 





' 1 

/ N\ 

1 N 


/ A\ 

/ 1 

X X X X 

X X 

.— > 

1 V 1 

1 1 

k 3 r ^ 

k o , 



N . 


' N 

/ A\ 

X X X X 

I V 

k 3 

/ , 

X X 

Now, the output of r-Deletion in (43) is of exactly the same 
structure as the /s/-irregular verbs where [-N] Redundancy Rule {'iH) 
applies (refer to the derivations in (35) and (36)). The rest of the 
derivation of the correct form ka:k-k'o is as follows: 






[-N] Red. 







Ass. (16) 



— > 

/ M 

X X X X 

1 1 

X X 


/ a\ 

X X X X 

/ 1 

X X 

1 V 1 

k a [t] 

1 1 

k o 

1 V 1 
k d k 

1 1 

Deletion of r in coda position creates an environment where [-N] Redundancy 
Rule (34) applies. The application of [-N] Redundancy Rule (34) is made 
possible, since r-Deletion (42) refers only to the segmental plane 
excluding the skeletal tier, and as a consequence, the x slot remains. 
Since the linear phonological representation wipes out the x slot as well 
as the segment r_ by rule (42), the assumption of the nonlinear phonological 
representation is essential. 

The alternation between r_ and t in the /t/-irregular verbs is thus 
accounted for by postulating /r/ as an underlying stem-final consonant and 
by making crucial reference to its position in the syllable structure. 

Before closing discussion of /t/-irregular verbs, the effect of adding 
/r/ to the Korean phoneme inventory needs to be discussed. Postulating /r/ 
as the stem-final consonant in /t/-irregular verbs greatly simplifies the 
grammar at the cost of increasing the number of the phonemes. Although the 
distribution of r_ and 1_ as allophones of the phoneme /I/ is completely 
predictable, it has been pointed out by Korean phonologists that r_ needs to 
be a phoneme. If /r/ is posited as an underlying phoneme, the grammar is 
more simplified, since /I/ is replaced with /r/ in the onset position, and 

hence the application of a rule, 1 — > r / V V, can be largely 

economized. Moreover, in the unmarked case, a language has nonlateral 
consonants as well if it has lateral consonant in the underlying phoneme 
inventory. It is also the case in Korean since all the underlying 
consonants are nonlateral except 1^. Therefore, it is anything but arbitrary 
to postulate /r/ in the underlying phoneme inventory. The phoneme /r/ is 
only constrained in its position in the syllable due to the surface 


constraint of Korean, 'N'. 

'4.3. /p/-Irregular Verb 

The 30-called /p/-irregular verb shows alternations as in (45), while 
the regular counterpart stays unaltered as already shown in (20) . 


















Formal . 

Ind. Conn. Stat. 

ko:p-t'a ko:p-k'o kow-a 

ku:p-t'a ku:p-k'o kuw-a 

mip-t'a mip-k'o tniw-a 


Ho be beautiful' 

^to roast' 

*to be hateful' 

The /p/-irregular verbs show a rather complex alternation between £ 
and w or u^. The analysis of this class of verbs follows Kim (1971), in 
assuming that the underlying representation of the stem final consonant for 
this class of verbs is /w/. Thus, the underlying representation of 
£-irregular verbs is like (46b). 

(46) a. regular verb 

k o p ^to be numb' 

b. irregular verb 


X X X X 

k o w to be beautiful' 

(46a) and (46b) are the underlying representations of the regular stem in 
(20a) and the irregular stem in (45a) respectively. 

What is to be accounted for in these data is the vowel length 
alternation and the w and £ alternation. Let's take, for example, the 
effective, formal, and stative forms in (45a). The lexical entries are as 

(47) a 

In the first cycle Syllable Building rules (4) build syllable structure. 
In the three representations in (4?) , (^Deletion (25) does not apply 
because the stem-final N' node is branching. Vowel Shortening (21) applies 


to (UTa.c) since the suffixes begin with vowel. In CJTb), the stem-final 
branching N' node blocks the application of ^-Deletion (26), and hence the 
structural description of Vowel Shortening is not met. Resyllabif ication 
applies at the end of each cycle. The verbal forms in 07) are represented 
as (48) when the lexical stratum is exhausted. 

(48) a. N" N" N" 

/ 1 < 

/N' /n' JH< 

/. ' I /I 
/ N ' N / N 

/ I / I / I 



k o w n i 

In the postlexical cycle, w in (48) is unspecified with respect to the 
categorial feature. However, it is underlyingly specified with the feature 
[round], which is to be incorporated into nucleus immediately before 
another nucleus by Nucleus Branching rule (49). 

(49) Nucleus Branching 



Nucleus Branching (49) creates a branching syllable nucleus, i.e. a 
diphthong. The structural description of Nucleus Branching (49) is met in 
(48a, c), but not in (48b). The newly-added nucleus node in (49) is 
automatically assigned the categorial feature [+N] by Nucleus [+N] 
Assignment (1). Thus, we derive the correct stative form kow-a by applying 
Nucleus Branching (49) to (48c). 

In (48a), however, the structure derived by Branching Nucleus (49) is 
N , which is possible neither as a long vowel (6b) nor as a 



diphthong (6c). Thus, the branching nucleus is reduced to a nonbranching 

one by deleting the unspecified nucleus x slot. 

(50) Monophthongization 


© > jg / X 



Monophthongization (50) spells out the nucleus x slot with the vowel u. 
Thus, we derive the correct effective form kou-ni by applying Nucleus 
Branching (U9) and Monophthongization (50) to OSa). However, the 
structural description of Monophthongization (49) is not met in the 
representation of the stative form in (48c) since the second x slot in the 
syllable head is underlyingly specified with a feature (I am assuming 
[♦low]), and the branching nucleus makes a well-formed diphthong structure. 

w in (ySb) is attached to the N' node and the categorial feature [-N] 
is assigned by Default [-N] Assignment (2). Thus, it is represented as 
[••■round] , which is phonetically realized as £. The application of 
Tensing TiS)* along with the phonetic realization of £ from [around] , 
derives the correct formal form ko;p-3'-»pni . 

In the derivation of the stratum I verbal forms except the formal 
form, the suffix-initial syllable nucleus is deleted not by the lexical 
rule (i>-Deletion (25) in stratum I, but by the postlexical rule 
Monophthongization (50). If the suffix-initial nucleus were deleted by 
(^Deletion (25), the structural description of Vowel Shortening (21) would 
not be met and the length of the stem vowel would not be affected by the 

In the following derivation (51) of the nominal form kou-m , the stem 
vowel becomes shortened by the suffix-initial nucleus which renains at the 
end of the lexical cycle. 


N" ^ 










1 i\\ 


X x X X 

X X 

— -> 

1 1/ 1 


k o w J 


Nucleus Branching (49) and Monophthongization (50) apply prior to [+N] 
Redundancy Rules (12) in the postlexical cycle, and the correct form kou-m 
is derived. This explains why the stem vowel length has shortened even 
though the suffix-initial vowel does not surface and the suffix begins with 
a consonant in the derived forms. 

The account of the alternations in the /p/-irregular verbs makes 
crucial use of underlying representation like (USb), underspecif ication of 
segmental features as well as categorial features, and their interaction 
with position in the syllable structure. 

Before closing this section, one more point needs to be discussed 
about the alternation between w or £ and £. The fact that the underlying 
/w/ alternates with either the vowel u or the consonant £ provides 
substantial evidence that the labial consonants and round vowels share some 
articulatory feature "representing constriction of the lips" (Walli 1984). 
In order to properly account for the phonological behavior represented by 
the alternations in /p/-irregular verbs, I suggest that the feature 


[labial] is more appropriate than [round]. 

H.y, /+/-Irregular Verb 

It has been noted in i.'i. that vowel-final stem verbs show regularity 
in their verbal paradigm. In the previous discussion of vowel-final verbs, 
verbs whose stems end with the vowel £ were omitted intentionally. This 
class of verb is regarded as one of the irregular types in Korean 
traditional grammar. In this section, I argue that +-final verbs are as 
regular as the other vowel-final regular verbs. Verbs in (52) belong to 
this class of verbs. 







yep'-- i-m 

yep'-- »-ni 




k'-- t-ni 





th— i-si 





yep'-J— ta 

yep'+- ko 










Propos. Formal, 

yep'— i-psi yep'-- J-pni 

k'-+psi k'-+pni 

th— J-psi th—J-pni 


^to be pretty' 
Ho extinguish' 
'to come out' 

When we compare the +-final verbs in (52) with other vowel-final verbs in 
(23), we find that what stigmatizes +-final verbs as irregular is the 
absence of stem-final vowel + before the stative suffix a. The stem of 
this class of verbs ends with the vowel £, since otherwise, +_ in the 
indicative and connective forms is not motivated. The absence of s in the 
formal form also supports the claim that the stems of the verbs in (52) are 
vowel-final. Thus, I suggest (53) as the underlying representation of the 
stem in (52a): 

(53) N 

y e 



Assuming the underlying representation (53). the derivation of the 
indicative and connective forms in (52) is straightforward since no rule 
applies but [+N] Redundancy Rules (12) to fill in the stem-final 
featureless x slot. Thus, yep'-i— ta and yep'4— ko are derived. 

In case of the verbal forms whose suffixes begin with unspecified 
nucleus, the structural description of Contraction (28) is met. The 
effective form yep'->ni In (52a), for instance, is underlyingly represented 
as (5ta): 

(54) a. 

N" N" 

N" N" 

I I 

N' N' 

I I 

N N 

I I 


I I 

n i 

b. N" N" N" 

Contr. /N' /N' /N' 




yep' £+)n i 


Contraction (28) is applicable to (.5^), since the two nucleus nodes across 
morpheme boundary are of the same representation (following the notational 
conventions in Pulleyblank (1983)) whether it is specified or unspecified 
with respect to the feature matrix. The remaining empty nucleus x slot is 
filled by [+N] Redundancy Rules (12), deriving the correct form yep'— »ni . 
Thus, deletion of [+] in the nominal, effective, honorific, and propositive 
forms is taken care of by Cksntraction (28), as seen in the discussion of 
vowel-final verbs. 

In case of the stative forms in (52), the stem-final vowel is deleted. 
The deletion of the stem-final vowel +^ is basically the same process as 
that of the suffix-initial vowel + in the vowel-final regular verbs shown 
in 3.4., in the sense that the unspecified nucleus x is deleted when 
abutting another nucleus x slot. 

We are thus in another position to provide evidence for the 
underspecified feature representation. (j>-Deletion (25) shows a potential 
for a mirror image rule since the unspecified nucleus x slot is deleted 
when abutting a specified nucleus x slot. That is, precise determination 
of vowel deletion is dependent upon the vowel quality, but not upon linear 
order in the vowel sequence. The evidence for the rule (25) being a mirror 
image rule is provided with respect to the stative forms in (52). 

The mirror image version of (^Deletion (25) 
(55) (^Deletion 

is (55); 


> J8 / 

Two adjacent syllable nuclei are not possible with an intervening morpheme 
boundary, if one of these nuclei is featureless. 

The derivation of the stative form yep'-a in (52a) is as follows: 

(56) a(r H" N"^ N" ^ 





In (56a) the last two syllable nuclei are adjacent. Since the stem-final 
nucleus is unspecified at the point of rule application, it is deleted. 

To avoid the clash of two nuclei, Korean phonology chooses to delete a 
syllable nucleus. The vowel to be deleted in a vowel sequence is 
determined by whether there is a nucleus which is not specified with 
respect to the feature matrix, as it is in the deletion of suffix-initial 
vowel -f. Within the framework of under specification, it does not cost for 


Q>-Deletion (25) to be a mirror image rule (55) since what counts is not the 
precedence of certain features but the presence or absence of feature 

The reason this class of verbs is marked as irregular is that the 
stem-final vowel ♦ is deleted, while no other vowels in that position are 
deleted as shown in vowel-final regular verbs in 3.4. However , ^-Deletion 
(25) and (55) which is motivated to account for the regular vowel-final 
verbs, predicts the deletion of the vowel + in the verbal forms in (52). 
What is involved in the stem-vowel deletion of the so-called /+/-irregular 
verbs is that their stem ends with an unspecified nucleus x slot in 
underlying representation, and this is exactly what theory of 
underspecification demands. 

It is explanatorily inadequate to deal with deletion of the vowel + in 
the same context as different processes: one as regular, the other as 
irregular. Rather, all the vowel-final stems follow the regular verbal 
paradigm: it is always the vowel [+] that is vulnerable in a vowel 
sequence. This phenomenon is accounted for in a simpler and more unified 
manner within the framework of underspecification. Without recourse to 
underspecification of features, the ^-deletion rule is more complex 
describing all the features for the vowels in a sequence, and still the 
phenomena of +- deletion is accounted for in an ad hoc manner (Lee 1976). 
Thus, the verbal forms whose stems end with a vowel provide substantial 
support for the theory of underspecification of vowel features. 

U.S. /1/-Irregular Verb 


/l/-irregular verbs show an alternation between 1^ and null. 

(57) Nom. Effect. Honor. Propos. Formal. 

a. k»:-m ka:-ni ka:-si k3:-psi ka:-pni 

b. mu-m mu-ni mu-si mu-psi mu-pni 

c. t+-m t+-ni t+- si t+- psi t4—pni 

In the Stratum II affixes, we see /I/ surfacing: 

(58) Ind. Conn. Stat. Gloss 

a. ka:l-ta k3:l-ko kar-a *to hang' 

b. mul-ta mul-ko mur-a ^to bite' 

c. t+l-ta t+l-ko t+r-a "to hold up' 

The stem-final consonant r in the stative is the alternant of 1^ in 
intervocalic position. 

We may hypothesize the underlying representations of the stems in (57) 
and (58) as: 

(59) a. N 

b. N 



X X X X 


1 V 1 

1 1 1 

k d 1 

m u 1 


For the stratum II suffixes, the only applicable rules are 
intervocalic /l/-weakening to r_ and Vowel Shortening (21). The verbal 
forms with the stratum I affixes in (57) systematically have no stem-final 
consonant 1^ present in the surface structure: I suggest a rule deleting /I/ 
in these forms. The stem-final 1_ is deleted before ra, ji, £, or s. That 
is, the deletion of the stem-final 1^ is not conditioned by a natural class. 
Thus, the environment of /l/-deletion makes crucial reference to the notion 
of stratum in Korean verbal morphology. 

(60) ^.-Deletion 

~> ® 

where w » (Domain: Stratum I) 

With this rule the stem-final 1^ is deleted, changing the thus empty x slot 
into nucleus x slot, whenever it is followed by a morpheme in stratum I. 
Since the stem-final x slot in the skeletal tier is assigned a syllable 
nucleus, the verbal stem behaves as if it were an +- final stem (compare 
with the verbal forms in (52)). Thus, 1^-Deletion (60) creates an 
environment for contraction (28). 

In discussion of /+/-irregular verb, it has been pointed out that 
there is an indeterminacy as to which vowel is to be deleted when both the 
stem-final and suffix-initial vowels are unspecified. We are now in a 
position to discuss the evidence that contraction rule (28) truncates 
stem-final vowel over suffix-initial one. The propositive form 
ka;-psi in (57a) is derived as follows: 



r N" 1 

Contr , 





1 1 \ 


1 V 

Ik » J 

p 3 i j 

In (61b) there are two unspecified nuclei across morpheme boundary, meeting 
the structural description of contraction (28). If the suffix-initial 
nucleus is deleted by contraction (28), then the unspecified stem-final 
nucleus can not be deleted after another specified nucleus because there is 


no morpheme boundary. Therefore, contraction rule (28) must delete the 
stem-final vowel, as shown in (61). 

The formal form ka:-pni in (57a) is derived as follows: 

Verbal forms in stratum II do not show any alternation except for the 
postlexical rule of weakening 1^ into r in intervocalic position. 

When we compare the verbal forms in /t/-irregular verbs in (39) with 
those in /l/-irregular verbs in (57) and (58), we find a striking 
difference in spite of the identical forms in the stative. Unlike the 
/t/-irregular verb, the /l/-irregular verb behaves as though the stem has a 
non-branching N' node. That is, /l/-irregular verbs follow the same verbal 
paradigms as /+/-irregular verbs when followed by stratum I verbal 
suffixes. The application of (^-Deletion (25), however, blocks the 
application of Vowel Shortening (21), since the morphemes in stratum I no 
longer begin with syllable nucleus at the point of the application of the 
rule (21). 1^-Deletion (60) applies in the beginning of stratum I, followed 
by ^-Deletion (26). The application of these two rules feeds contraction 
(28) and ©-Deletion (25). 

The 1^-Deletion (60) before certain morphemes in /l/-irregular verbs 
can be accounted for only by appealing to the notion of stratum. Basically, 
it is also this rule that determines the applicability of contraction (28), 
(deletion (25), and Vowel Shortening (21). 

4.6. /1+/-Irregular Verb 

/l+/-irregular verbs are labeled so because 11_ appears instead of 1+ 
in the stative form. In the other verbal forms, the phoneme /I/ in 1+ 
weakens to r since it is in intervocalic position. 








Formal . 



kar— t-ni 

kar— fsi 

k»r— J-psi 

kar— 1-pni 


mur— i-m 

mur— J-ni 

mur— t-si 

mur— J-psi 

mur— J-pni 


kur— J-m 


kur— fsi 


kur— I-pni 









'to filter' 


mur-»— ta 

mur-i— ko 


'to be soft' 


kur-I— ta 

kurV- ko 


'to roll' 

The underlying representation of the stem in (63a) is (6'<), 

(64) N 

X X X X 

k 3 1 1 

With the underlying representations like (64), /l+/-irregular verbs 
follow the same paradigm as /l/-irregular verbs. The formal form kdr— »pni 
in (63a) is derived as follows: 


X X X X 

I M I 
k a 1 [+]p n i 

In (65b), stem-final 1^ is deleted and the empty x slot becomes syllable 
nucleus by 1^-Deletion (60) since it is followed by stratum I suffix. As a 
consequence, the verbal stem behaves like a verb whose stem ends with vowel 
+, and ^-Deletion (26) applies. By deleting s, two identical syllable 
nuclei are left adjacent across morpheme boundary, as shown in (65c). The 
application of contraction (28) derives the corret form kar-i-pnl . 

In the derivation of a formal form in (65), it has been shown that 
/l+/-irregular verbs follow the same derivational processes as 
/l/-irregular verbs (compare the derivation in (65) with the one in (62)). 
The account of the /l+/-irregular verbal forms makes use of rules necessary 
to account for /I/- or /-J-Z-irregular verbs, assuming the underlying 


representation like (64). That is, both (p-Deletion (25) and 1^-Deletion 
(60) are motivated independently of /l+/-irregular verbs. 

In case of stratum II verbal forms, ^-Deletion (60) does not apply, 
just as it does not in /l/-irregular verbs, since the domain of 1^-Deletion 
(60) is restricted to stratum I. Thus, the derivations of the stative and 
connective forms kgll-a and kar-t— ko are as follows: 

(66) a. 

X X X X 

I I I I 

lU 5 1 1 j 

N" ^ 




X X X X 

I I I I 

L^k a 1 1 J 

k o 

In (66a, b), no rule is applicable in the second cycle and resyllabification 
applies at the end of the cycle. Thus, the stative form kgll-a is derived 
in (66a) with no more rule to apply at the postlexical cycle. 

In case of (66b), however, resyllabification applies at the end of the 
lexical cycle and leaves an unsyllabified x slot linked to 1^. In the 
postlexical cycle the unsyllabified x slot linked to 1^ is syllabified as a 
nucleus node losing its feature 1^ . 

(67) Nucleus Assignment 

By assigning nucleus node to the unsyllabified x slot the unsyllabified x 
slot is syllabified. 

The unsyllabified x slot at the end of the second cycle in (66b) is 
syllabified as nucleus of an empty x slot. Nucleus Assignment (67) derives 
(68) from (66b). 

[+N] Redundancy Rules (12) fill in the derived empty x slot in (68) 
and resyllabification applies at the end of the postlexical cycle. 


Intervocalic 1^ becomes £ and the correct form kar-i— ko is derived. 

In spite of the contribution that the underlying representation like 
(6'4) for the /l+/-irregular verbal stem makes toward a simplified grammar, 
the underlying representation in (6'4) remains problematic: the 
representation (6'4) is less highly valued according to the McCarthy's 
(1981) revised version of Obligatory Contour Principle (OCP), since there 
are two adjacent x slots of identical segment 1_. On the other hand, if the 
underlying representation of /l+/-irregular verbs are like (59), meeting 

(69) N 

k a 1 

then, there is a difficulty in applying ^-Deletion (60) and Nucleus 
Assignment (67). These rules allow only one half of the geminate structure 
to undergo the structural change, which is strongly rejected by 
phonologists: Kenstowicz (1982) defends geminate structure constraint at 
the cost of reducing the OCP to a language specific parameter. Steriade 
and Schein (igSU) make the geminate structure constraint even stronger in 
case the geminate structure itself is the "target", rather than the 
"context", of the rule. Following their assumption that the geminate 
structure constraint is not to be violated, the present analysis chooses 
without further language specific evidence the underlying representation 
like (62) violating the OCP, over the underlying representation like (67) 
violating the geminate structure constraint. 

To sunmarize the discussion of the /l+/-irregular verbs, they behave 
like /+/-irregular verbs in stratum I due to the application of 1^-Deletion 
(60) as /l/-irregular verbs do. It has already been argued that 
/♦/-irregular verbs are as regular as other vowel final verbs. Therefore, 
it follows that neither/1/- nor /l+/-irregular verbs are irregular. The 
regularity of the /l+/-irregular verbs are accounted for with reference to 
the nonlinear underlying representation like (64), the syllable structure, 
and the lexical stratum. 


It has been argued that the so-called irregular verbs are different 
from the regular verbs in their underlying representations, and that as a 
consequence, the structural descriptions of rules are not met in the 
irregular verbs though they are met in the regular verbs. This line of 
deeper phonological distinction is made possible by nonlinear phonological 
representations . 

In addition to the rules needed to account for the alternations in the 
regular verbs, the so-called irregular verbs require the following rules: 
as a lexical rule, 1^-Deletion (60) applies in stratum I. The mirror image 
version (55) of g) -Deletion (25) also applies in the lexical cycle. Among 
the postlexical rules are [-N] Redundancy Rule (34), Glide Formation (37), 
r-Deletion (42), Nucleus Branching (49), Monophthongization (50), and 

Nucleus Assignment (67). 

Lexical rules of 1^-Deletion (60) and ^-Deletion (26) which are 
specified with respect to the domain of rule application must apply in the 
beginning of each cycle in stratum I. Of those two rules, 1^-Deletion (60) 
applies first, feeding s-Deletion (26). These two rules feed contraction 
(28) and (^Deletion (25), blocking the application of Vowel Shortening 

When the lexical cycle is exhausted, postlexical rules apply. They 
apply as soon as their structural descriptions are met. Thus, no extrinsic 
ordering is necessary. For example, ^-Deletion (42) creates the 
environment for [-N] Redundancy Rule (3'4), and Nucleus Assignment (67) 
feeds L+N] Redundancy Rules (12). When the structural descriptions for 
both Tensing (13) and Assimilation (16) are met. Tensing (13) must apply 
prior to Assimilation (16) so that the one feeds the other. The reverse 
ordering bleeds Tensing (13) since the application of Assimilation (16) 
creates a geminate structure and the application of Tensing (13) violates 
the geminate structure constraint, allowing only one half of the geminate 
structure to undergo the structural change. Resyllabification applies in 
the postlexical as well as lexical cycles, since it applies at the end of 
each cycle. 

The two redundancy rules (12) and (34) are intermingled to interact 
with other phonological rules. The two rules Nucleus [+N] Assignment (1) 
and Default [-N] Assignment (2), which are concerned with the categorial 
features [+N] and [-N] , also interact with redundancy rules on the one 
hand, and with other phonological rules such as Nucleus Branching (49) and 
Monophthongization (50) on the other. 

5. Conclusion 

The nonlinear analysis of Korean irregular verbs illuminates a 
phonological regularity in the verbal system which was obscured in previous 
linear accounts. The allocation of the surface phonetic contents into 
separate planes gives room for differentiating the regular and irregular 
verbs in their underlying representation. Since the phonological rules in 
the framework of nonlinear representation make reference to syllable 
structure, the skeletal tier, and the lexical strata, in addition to the 
underlying phoneme, the regularity of the irregular verbs as well as the 
regular ones is accounted for. Therefore, in nonlinear analysis, two 
different types of verbs are differentiated by the underlying 
representation without appealing to lexical exceptions to rules or using 
other non-phonological diacritics. 

From the theoretical point of view, the present analysis supports 
several recent theoretical issues. First, the idea of featureless x slot 
in the skeletal tier in Clements and Keyser (1983) is exaraplified in the 
behavior of the /s/-irregular verb. The alternations in this class of 
verbs are best treated by assuming a stem-final empty x slot. 

Secondly, the theory of underspecification allows rules to be 
formulated in a simpler way. The notion of specifying only marked features 
in the underlying representation whether the value is + or -, and the 


interaction between the redundancy rules for the unspecified slot and the 
other phonological rules, provide a principled way of accounting for the 
deletion of the vowel ♦. 

A ramification of underspecification of features accounts for why the 
diphthongs •w* and *^ are systematically excluded in Korean. Since the 
unspecified x slot in the branching nucleus node is deleted by 
Monophthongization (50), these structures cannot surface as diphthongs. 

Thirdly, the notion of lexical stratum is crucial in accounting for 
the regular phonological processes at a deeper level. Without the notion 
of stratum, the alternation before certain classes of suffixes is viewed to 
be conditioned by segments which do not constitute a natural class, and 
hence, becomes explanatorily inadequate. 

Finally, I suggest that the feature [labial] is more appropriate than 
[round] in capturing the phonological altenation between the round vowel 
and the labial consonant as well as in defining the articulatory 
characteristics . 


For the preparation of this paper I was greatly helped by Professor 
Diana Archangeli with her insightful comments and suggestions through 
marathon discussions. I thank Professor C.-W. Kim for his critical 
comments. I would also like to express my appreciation to Jean D'souza for 
helping me to crystallize and refine the ideas herein. I am, however, 
solely responsible for any problems which remain. 

The Korean orthography, which used to be regarded as representing 
the underlying phoneme, has been used to name the irregular verbs. 
Although this traditonal labeling may be misguided, the labels in this 
paper follow it. 


Mohanan (1981) allows for a "loop". That is, stratum n feeds into 

stratum n-fl as usual, but it also feeds into stratum n-1. 

Neutralization in Korean also involves the neutralization of 
aspiration and tense into unmarked features in syllable final position. 
However, this process is not crucial to the present analysis, and it is not 
specified in Neutralization (14). 


It can be claimed in general that the least marked vowel is more 

likely to be truncated or inserted in the hierarchies of deletion and 

epenthesis processes. For further discussion, see Archangeli (1984) and 

Sohn (to appear). 

The underlying representation of the stative suffix is /»/ . The 

representation of a for the stative suffix is mainly used for typographical 

convenience. However, /»/ becomes a by Vowel Harmony when the stem-final 

vowel is a or o. We do not go into discussion of Vowel Harmony since it is 

not relevant in this paper. 


There is indeterminacy at this point as to which one of the two 
identical slots is to be deleted by contraction (28): the stem-final 
or the suffix-initial nucleus. Preference to the first is argued in the 
derivation of (61) in /l/-irregular verbs. 


The traditional Korean orthogrphic system, which is generally 
regarded as a reflex of Korean phonemes, dictates that the stem of /s/- 
irregular verbs be represented with £ when it is followed by 
consonant-initial suffix. But s is not present orthographically when 
followed by vowel-initial suffixes. This is why this class of verb is 
called irregular. 


The alternative way of filling in the empty x slot in coda position 

is to invoke Consonant Spreading from right to left: 

N' N" 
I I 

X X 

[ ] 
The preference of [-N] Redundancy Rule (S'O over Consonant Spreading is 
motivated by the resultant structure after the empty x slot is filled by 
spreading. Later application of Tensing (13) or Assimilation (16), which 
must be made in the derivation of this class of verbs, violates the 
geminate structure constraint (McCarthy 1981) by allowing only one half 
of the geminate structure to undergo the structural change. 


I suppose that [-N] Redundancy Rule O^*) is a more general rule in 
Korean, accounting for the tensification phenomenon where an obstruent is 
tensed without any preceding tense-triggering consonant. Recall the general 
Tensing (13): an obstruent is tensed when preceded by another obstruent. 
/n«*-ka/ "riverside' is realized as /nstk'a/ with a tensed obstruent. 
If the tensification in compounding is reduced to insertion of an empty x 
slot between the first and second morphemes, then the empty x slot in coda 
is filled by [-N] Redundancy Rule (34) triggering tensing of the following 

XX XX — > X x[x]x X — > X X X X X > X X X X X 

natka natka n stitjk a n ie 1< k'a 

Tensing (13) and Assimilation (16) apply in the last step of of derivation. 

The feature matrices in Glide Formation (37) are not described in 
terms of underspecification of features because the whole picture of 
under specification of vowels is not elaborated at this point. Roughly 
speaking, the vowel i^ is specified with [-back], and u with [+round]. 
Unlike the SPE theory in which i^ and u constitute a natural class 
represented as [+high, +sonorant], underspecification of features does not 
allow for the natural class of these two vowels. However, these two vowels 
are differentiated from other vowels in that they are not specified with 
categorial feature in the underlying representation. They are 
differentiated from consonants in that they are possible syllable nuclei. 

Thus, Glide Formation (37) can be reformulated in terms of 
underspecification as follows: 

(37)' Revised Glide Formation 

[ ] [ ] 



The empty bracket [ ] refers to a specified feature matrix attached to x 
slot. With this formulation, Glide Formation (37)' must apply prior to 
Nucleus [+N] Assignment (1) since it crucially refers to whether the 
specified feature under N node is specified with the categorial feature. 

In Middle Korean the irregular verb cis ^to make' has /s/ as a stem 
final phoneme. Historically, the voiceless fricative £ has either changed 
to voiced z which eventually disappears, or remained unchanged. 

The /s/-irregular verbs in (30) follow the former type of historical 
change when followed by vowel-initial suffix, while they follow the latter 
type when followed by consonant-initial suffix. Therefore, assuming /cis/ 
or /ciz/ as the underlying representation of cis does not solve the problem 
of explaining two alternations that a single underlying phoneme shows. 
Moreover, it is explanatorily inadequate in the synchronic grammar since 
there is a class of verbs as in (10) whose stem-final £ does not become 
silent before vowel- initial suffix. 

Furthermore, it is not likely that a child acquires a language on the 
basis of the historical model. Therefore, there is no /s/ present in the 
underlying representation of this class of verbs. 


I follow Kim (1971) in assuming that the stem-final consonant of 

the /t/-irregular verb is /r/. 

^-Deletion (42) was suggested to me by Professor Diana Archangeli, 
whose insight captured the similar processes going on in the derivations of 
/s/- and /t/-irregular verbs in Korean. 


Nucleus Branching is independently motivated with respect to the 

syllable structure of Korean. As in Spanish (Harris 1983), glides ^ and w 

in Korean are not onset but parts of nuclei. For further argument, see 

Sohn (to appear) . 

As mentioned in note 6, there still remains indeterminacy on 
deleting a vowel by contraction (28) . 

Refer to note 4. 


/l/-irregular verbs, unlike the already discussed irregular verbs, 

do not have regular counterparts. 


1^-Deletion is an independently motivated phenomenon when 1^ is a 

morpneme-final consonant and it is followed by another morpheme. First, 1^ 

in the reduplicated morpheme is deleted. Some adverbs in Korean are 

derived by reduplication of a noun, i^ in the following two examples is an 

adverbial marker. 

a. tal ^roonth' /tal+tal+i/ — > ta-tal-i ^every month' 

b. nal "day' /nal+nal+i/ — > na-nal-i "every day' 
Secondly, Verbal C>>mpoundings show 1^-deletion, 

c. y>l "to open' /y»l+tat/ — > y*-tat "to open and shut' 

d. mil "to push' /mil+tat/ — > mi-tat "to slide-open* 
Thirdly, noun compoundings also show 1^-deletion. 

e. sol "pine' + namu "tree' — > so-namu "a pine tree' 

f. pul "fire' ♦ napi "butterfly' ~> pu-napi "a tiger moth' 

g. pul "fire' + sap "shovel' — > pu-sap "a fire shovel' 


According to 1-Deletion (60), 1 in these examples becomes +, which will 
eventually derive incorrect forms. + needs to be deleted. 

In order to account for the phenomena above or /l/-irregular verbs the 

linear rule 1 — > j0 / ]y (where v t 0) is sufficient. However, 1^-Deletion 

formulated as (60) sheds light on the phonological alternation in 
/l+/-irregular verbs. 

^^ The general way of taking care of unsyllabified x slot in Korean is 
deleting it, rather than epenthesizing a vowel. For example, 
a. b. 

Tn" 1 





X X 


i a J 

-"N" 1 





X X 

lU 'i t J 

t a J 


stop and the preceding 
is delinked and the 


(a) and (b) above are derived as ap-ta Hhere is nothing' and an-ta 
sit' respectively. 

However, when the unsyllabified x slot is 
syllabified x slot is not, the continuant 
noncontinuant x slot is incorporated into 

c. [-continuant] Incorporation 

I ^ '-^ 
t+cont] [-cont] 
The derivations of pa;p-ta *to tread' 

d. e. 


-ta Ho be bright' are as 


/V\ )t X X 

p^ipta pail<ta 

The unsyllabified x slot is syllabified delinking the already syllabified x 
slot, when the unsyllabified x slot is [-continuant]. Nucleus Assignment 
(67), however, is not applicable to the resultant unsyllabified 1_ in (d) or 
(e) because it violates the Well-Formedness Condition that association 
lines do not cross. Thus, Nucleus Assignment (67) is not contradictory to 
other syllabification processes in Korean. 


ARCHANGELI, Diana. 198M. Underspecification in Yawelmani phonology and 

morphology. Ph.D. Dissertation. MIT. 
BAKER, M. 1983. Some concepts and consequences of a theory of diphthong 

representation. MIT ms. 
CHOMSKY, Noam and Morris Halle. 1968. TTie sound pattern of English. 

Newyork, Evanston, and London: Harper and Row. 
CLEMENTS, G.N. and S.J. Keyser. 1983. CV phonology :a generative theory of 

syllable. Linguistic Inquiry Monograph Nine. 
HARRIS, James. 1983. Syllable structure and stress in Spanish: a nonlinear 

analysis. Linguistic Inquiry Monograph Eight. 


KENSTOWICZ, Michael. 1982. Gemination and Spirantization in Tigrinia. 

Studies in Linguistic Sciences 12:1. 103-122. 
KIM, C-W. 1971. Regularity of the so-called irregular verbs in Korean. In 

C. W. Kisseberth, ed . : Studies in Generative Phonology, pp. lOU-IIS. 

Edmonton, Canada: Linguistic Research Inc. 
KIM-RENAUD, Young-Key. 1973. "Irregular" verbs in Korean revisited. 

Language Research 9.2:206-225. 
KIPARSKY, Paul. 1982. Lexical morphology and phonology. In Linguistic 

Society of Korea, ed.: Linguistics in the Morning Calm, pp. 3-91. 

Seoul, Korea: Hanshin Publish Co. 

. 1983. Some consequences of lexical phonology. MIT ms. 

. 1984. On the lexical phonology of Icelandic. MIT ms. 

LEE, P-K. 1976. Underlying segments in Korean phonology. Ph.D. 

Dissertation. Indiana University. 
LEVIN, Juliette. 198'4. Conditions on syllable structure and categories in 

Klamath phonology. 
MASCARO, Joan. 1978. Catalan phonology and the phonological cycle. Ph.D. 
Dissertation. MIT. reproduced in Indiana University Linguistic Club. 
McCarthy, J. I98I. a prosodlc theory of nonconcatenative phonology. 

Linguistic Inquiry 12. 373-'*18. 
MOHANAN, K.P. 1981. Lexical phonology. Ph.D. Dissertation. MIT. 
SOHN, Hyang-Sook. to appear. On the representation of vowels and diphthongs 

in Korean . 
STERIADE, Donca. 1982. Greek prosodies and the nature of syllabification. 

Ph.D. Dissertation. MIT. 

. and Barry Schein. ^S&^. On geminates. MIT ms. 

WALLI, B. 1984. On the representation of complex segments and their 

formation in Kinyarwanda. MIT ms. 

Studies In the Linguistic Sciences 
Volume 15, Number 1, Spring 198 5 

Sex, Power and Linguistic Strategies in the Hindi Language 

Tamara Valentine 

This paper explores the complex relationship among 
power, gender, and linguistic strategies in the 
non-Western, Indo-Aryan language, Hindi. Examining 
data from contemporary literary text-based Hindi 
cross-sex conversations (Ashk 1976, Rakesh 1958, Verma 
I96M), evidence is provided that a differential use of 
verbal interactional patterns is held by male and 
female Hindi speakers, and that these linguistic 
strategies are verbal expressions of strength or energy 
which are interpreted and motivated differently by each 
sex for the purposes of coordinating interaction and 
achieving effective communication with each other. 
Specifically, the discoursal strategies of 
conversational topic Initiation, maintenance, and 
shift, gaining an appropriate response and establishing 
a common theme, and the organizational devices used to 
regulate and coordinate discussion are excunined. It is 
revealed that the female Hindi speakers are primarily 
the maintainers and supporters of cross-sex 
conversation and the male Hindi speakers are the 
relators of events, conveyors of messages, and gainers 
of information. 

1.0 Over the past few years a considerable number of studies 
has emerged on the sexual differentiation of language and speech, 
shedding light on the complex, varied ways in which males and 
females verbally interact with each other. The primary data and 
research on cross-sex interaction come from actual conversations of 
American couples (Fishman I98O, 1983; Leet-Pellegrlni 198O; Sattel 
1983; West and Zimmerman 1983; Zimmerman and West 1975). It is 
shown that when women and men verbally interact as equals or 
intimates they do not necessarily make use of the same discoursal 
skills, nor do they play the same roles in cross-sex conversations. 
By the examination of male-female discourse, gender differences in 
such strategies as the patterns of topic initiation, interruption, 
question asking, and conversational style, to name a few, emerge. 
The explanation researchers offer for these verbal interactional 
differences between the sexes is found in the interdependent 
relationship among language, gender and power: the hierarchical 
relations in cross-sex discourse are attributed to male-female power 


relations In society. In view of the fact that the norms of 
appropriate behavior in society establish and enforce the power and 
control for men and are withheld from women, it is assumed that 
males also control, dominate, and command power in cross-sex talk, 
"Just as male dominance is exhibited through male control of 
macro-institutions in society, it is also exhibited through control 
of at least a part of one micro-institution" (Zimmerman and West 
1975:125). Hence, power is seen as a measure of dominance which is 
exercised by one language user over another. A powerful speaker is 
one who directly and indirectly influences others through discourse. 

As an alternative to viewing power as a restriction of freedom 
and as the ability to control, dominate, and determine the goals and 
behaviors of another, Hartsock (1981) suggests that power in 
language discourse is 'energy, effective interaction, and 
empowerment.' In verbal interactions "no compulsion is present 
other than the force of discourse itself; domination is absent, and 
reciprocity pertains between and among participants" (Elshtain 
1982:620). In this sense power is not understood as a system of 
dualism: powerful vs. powerless; dominating vs. dominated; and 
winner vs. loser, but is presented as a strategy that affirms the 
patterns found in females' and males' speech. Each speeiker's act In 
itself is an act of strength and energy. Power is effective 
communication, an accomplishment understood to be satisfying in 

Talking into account the view of power as energy, I examine its 
relationship to language strategies in cross-sex conversation of the 
non-Western language, Hindi. Such an exploration reveals how 
discoursal work is done in cross-sex Hindi conversation, provides 
evidence that different discoursal models are held by male and 
female Hindi speakers, and suggests that these linguistic strategies 
are acts of power used for the purpose of attaining and achieving 
efffective communication. I provide detailed, illustrative examples 
from literary text-based cross-sex conversations in the Hindi 
language. Primarily data are drawn from three contemporary Hindi 
works: one novel and two plays of well-known contemporary authors of 
India (Verma 1964; Ashk 1976; Rakesh 1958, respectively), eunong 
other Hindi short stories to substantiate my claims. Each creative 
text contains conversations between intimate, heterosexual Hindi 
speaking couples. I isolate what verbal strategies and skills are 
used and examine how these particular linguistic patterns are 
employed by male and female Hindi speakers in order to communicate 
with each other. Specifically, I examine the strategies of 
discourse topic initiation and the success rate in gaining an 
appropriate response and in establishing a common theme, how this 
discourse topic is maintained or shifted, and the organizational 
devices used to regulate verbal contact and conversation. Analysis 
reveals that these discoursal strategies function to ensure, 


encourage, and direct conversation between the couples. The 
different use of these strategies by the Hindi speakers does not 
suggest that inequality of power exists in the verbal interaction 
between the sexes or is due to one of the parties exerting control 
over the other, but that each of the speakers exercise her or his 
power in conversation differently as a drive to interact 
effectively. I am making no claim that these patterns of cross-sex 
interaction are generalizable to live conversations in Hindi; no 
empirical investigation of natural actual speech has been done as 

2.0 One aspect of effective communication within the Hindi 
dlscoursal system of turn-taking is in Initiation and maintenance of 
conversational topic. Evidence for the strength of topic initiation 
has been presented in many linguistic studies by analyzing the 
success rates of raising, adopting, and maintaining a topic in an 
on-going conversation (Coulthard 1977; Keenan and Schieffelin 1976; 
Sachs, 197'*; Soskin and John I963, as well as the 
aforementioned research). An utterance is considered an initiation 
of topic when the subject raised is different from that of the 
preceding utterance. However, initiating a topic is only an attempt 
to get a discussion started, it does not necessarily guarantee that 
it is developed nor ensure that the theme is adopted and maintained 
in conversation. In order for a discourse topic to evolve into an 
actual discussion and be considered successful, both of the speakers 
must work and contribute to the topic's development. Not only must 
one participant raise the topic, but the other must respond, thus 
mutual orientation and direction is displayed to each other and to 
the topic at hand. Although topic initiation is a strategy 
available to both speakers in a cross-sex conversation, it is not 
motivated, understood, and interpreted similarly and the success 
rates differ. 

With a possible gender differential usage of topic initiation 
and maintenance in mind, the following questions come into focus: 1, 
Does the female or the male Hindi speaker initiate discourse topic 
more often? 2. Who is more successful in gaining an appropriate 
response? and 3- Are there statistical and motivational 
differences in initiating discourse topic with a question or 
statement form? Firstly, from the cross-sex Hindi conversations 
examined a count of 118 discourse topics are raised. Of this, 601t 
are introduced by the male speakers and >iO% by the female speakers. 
The male speaker is 83% successful in initiating topic, while the 
female speaker is 60% successful. It is evident that there is a 
clear difference between the sexes in successful initiation of 
conversational topic in these texts - the male speaker Initiates 
topic more often and is more successful in gaining a response from 
his listener than the female speaker. Secondly, in these same 
creative Hindi works the male speaker raises conversational topic 


with a statement 64^ of the time and the female speaker 52%; the 
male raises conversational topic with a statement 36/t of the time 
and the female 4851. The male speaker is 80% successful in 
Initiating with a statement or a question, while the female speaker 
is 53% successful with a statement and 60% with a question. In 
other words, the following pattern emerges: although both sexes use 
the option of presenting discourse topic with a statement or a 
question the male Hindi speaker predominantly initiates with a 
statement, and the female Hindi speaker initiates with a question 
slightly more frequently than with a statement. 

Although both question and statement forms are opportunities to 
initiate discourse topic there is a clear difference in speaker 
motivation in the varied use of them by each sex. Generally, it is 
argued that questions are stronger forms interactively in that they 
are requests for Information, action, and acknowledgments and demand 
an answer, whereas declarative statements are more easily ignored 
and less verbally recognized because they provide information and do 
not require a response. If question asking is viewed as a strategy 
for the goal of maintaining conversation and for sustaining the 
topic and not as a procurement or an elicitation of an answer or 
reaction, then the generalization that questions are stronger forms 
than statements does not necessarily hold. In the Hindi cross-sex 
passages to follow, it will be seen that for the female speaker 
effective communication lies in the power of tying together, filling 
in, and linking utterances and topics to support and create 
continuity in conversation. Therefore, for the female Hindi 
speaker, questions serve to maintain an uninterrupted, friendly 
stream of talk. For the male Hindi speaker, however, questions are 
used primarily to attract the attention of his listener for the 
purpose of requesting information. Statement forms, on the other 
hand, express facts, rules, attitudes, explanations, and 
descriptions. The male Hindi speaker tends to initiate discourse 
topic with a statement for the reason that this form suffices to 
elict the attention of his listener in order for the maintenance of 
his conversational topic. The female Hindi speaker, however, uses a 
statement to initiate discourse topic as a conversational strategy 
of maintaining continued talk, not as an utterance that ensures the 
success of her turn. 

Passage I below, from Mohan Rakesh's (1958:45-48) play Ashafh 
ka ek din. Illustrates that the use of question forms by the female 
speaker is one discoursal system of sustaining a natural flow of 
topic and conversation. The use of questions and tags by the female 
participant, Mallika (M), in lines 1,3, and 5 is not for the 
purpose of seeking a verbal response or even a verbal acknowledgment 
from her male partner, Kalldasa (K). Rather these forms function as 
a strategy of elicitation of attentiveness from and contact with her 
listener as well as of continuity and maintenance of the 


conversation in general. No replies are required or expected. On 
the other hand, In lines 7 and 9, the question forms which Kalldasa 
asks, function to gain a response and are viewed as requests for 
Information. He is soliciting permission from Mallika to leave the 
village. Mallika interprets Kalldasa's questions as requests or 
inquiries and replies appropriately. The female participant 
provides the necessary acknowledgments (e.g., ha and nahf) and 
complements her responses with the use of statements, which are 
expressed in the forms of internal reports about her feelings and 
her future plans. In spite of the fact that these utterances 
provide new information, they are not for the purpose of gaining a 
successful turn but for maintaining continuance of the verbal 
interaction with her male partner. 


1. M: phir udas ho gave? Idekhoj tun 
Are vou brooding again ? JLook, 

mujhe vacan de cuke ho. 
you promised me. 

turn phir ek bar soco Mallikal prashga saraman 

Dr rajyaghray_svlkar karne ka hi nsihl h£. 

usse kahi bara prashna mere samne ht. 

Think again Mallika ! It isn't only a question of 

accepting the honor and patronage. I have a bigger 

question than that^ ^ ^. 

M: 3r vo prashna m£ hu hu na . yaha bttho. tum mujhe 

jante_ho. Ji£^aa? tum samajhte ho ki turn is avsar ko 
thukrakar vaha rah jaoge to mujhe sukh hog a? 
|mfe lantl hu| ki tumhare cale lane _ 
par mere antar ko ek riktta cha legi, 3r bahar 
bhi sambhavatah bahut^suna pratlt hoga. phir bhi ^ 
mt apne sath chal nahl kar rahl. rat hrday se kahtl hu 
ki tumhe jana cahiye. 

And I am the question, aren't I? Sit here. You know me, 
don't you? You think that by turning down this 
opportunity and staying here you will make me happy f 
|l know] that when you go, I will feel empty 
inside, and outside I'll feel very lonely too. 
Even then, I'm not deceiving myself. From the 
heart I^say that you must go. ^ ^ 

K: cahta hu ki tum is samay apnT akhe dekh saktT. 
I wis^ you could see your eyes right now. ^ 

M: merl akhe isllye gill he ki tum raerl bat nahT samajhte. 
tum vaha se lakar bhT mulhse dur ho sakte ho? . . . yaha 
gram-prantar jne rahkar tumharl pratibha ko viksit hone 
ka avkash kahi milega ? yaha log tumhe samajh nahT pate 
ht. ve sam3nya kT kasstl par hi tumharl 
parTksa^ka rna cahte h T. vishvas karte ho na ki mc tumhe 
lantr hu ? Ilantr hul ki koT bhT rekhS tumhe gher le 
to turn ghir jaoge. mT tumhe^herna 
nahl cahtT. isTliye kahtT hu ki tum jao. 
I have tears in my eyes because you don't seem to 
understand what I'm saying. Even though you go away from 
here will you be far from me? If you stayed in the 


village, would vou ever find the time to develop 

vour genius? These people here can't understand 

you. They can Judge your ability only on general 

criteri a . Don't vou believe that I understand vou? 

Il know| that if any line encircles you, you will be caught. 

I don't want to hold you. Therefore I say, 

that you should go. ^ _ , -, 

6. K: turn mujhe purT tarah nahT saraajh r;ahi ho |malikkal 

prashna tumhare gherne ka bhi nahl h£-. . . . 

You don't understand me completely i Mallika l It's not a 

question of you holding me 

7. K: mu.ihe lane ke live kah rahT ho ? 

Are you telling me to go ? 

8. M: Iha^ . jdekhna l, me tumhare pTche prasann rahugl, bahut 

ghumugT or har sandhya ko Jagdamba ke mandir me suryast 
dekh n e iava karuej . . . 

[Yesl IYou'11 see[ I'll be happy after you leave. I'll 
wander and watch the sunsg,t from the temple.... 
" hf ':uiRi5e \ ,^,_ 
^i^. ,;i^^ meg.. 1 mu-it leave vou ? 
10. A: I nahiflj vida tumhe naht d'ugl. ja rahe ho, isliye keval 
prarthna karugT ki tumhara path prashast ho...jao. 
[^oi I'm not telling you to leave. You are going so I 
only pray that your path is clear... go. 

I suggest, then, that rather than positing that the male Hindi 
speakers are claiming conversational control and are primary holders 
of the floor and attributing the differential characteristics of 
initiation of discourse topic to the male dominance and power 
structure in Indian society, I treat this linguistic disparity in 
initiation of topic between the sexes simply as one discoursal 
strategy of power motivated and used by two speakers for the mutual 
attainment of effective cross-sex communication. Each strategy 
which each speaker empowers is in itself an act of strength, and it 
is not that one speaker has more control (or is more powerful) than 
another only because one participant initiates more discourse topics 
or one uses less question forms. Rather than assuming that such 
strategies as initiation of topic are a means of procuring power and 
control, they are understood in terras of neutral contributions to 
the conversation for better communication. The differential use of 
this discoursal strategy, then, is explained by suggesting that male 
and female Hindi speakers possess different interpretations of 
question and statement forms in initiating topic and sustaining 
conversation. For the female Hindi speaker, both questions and 
statements are verbal strategies used to maintain and support 
conversation. For the male Hindi speaker, question forms are 
motivated as acts of requests for information, and statements are 
acts of elicitation for attention to ensure a successful topic 


3.0 A second aspect of effective communication in Hindi 
cross-conversations which varies with gender are the linguistic 
conventions of topic flow and topic shift. Once the conversational 
topic has been introduced, it evolves and changes differently for 
each sex. For the female Hindi speaker, the system of developing 
discourse topic progresses gradually and flows smoothly; shifting 
from one discourse topic to another is usually not sudden or 
unexpected. Acknowledgments which recognize and evaluate the 
preceding utterance are made and development of the topic is 
pursued. For a male Hindi speaker, on the other hand, 
conversational topic tends to be narrowly defined and topic shifts 
tend to be disconnected. 

Passage II below, an excerpt from Upendranath Ashk's 
(1976:78-79) short story tpliye , illustrates the varied linguistic 
use of topic flow and shift by male and female Hindi speakers. The 
female participant Madhu (M) and the male participant Vasant (V) are 
discussing the subject of appropriate behavior. A new subject, the 
participants' "cousins Ushi and Nimmo", is introduced by the male 
speaker (line 11); the preceding utterance has no reference to this 
subject matter. The female speaker recognizes and announces th^s 
abrupt shift in topic and exclaims " ipne phir ushT jt nimmo kl bat 
cheri 'again you bring up Ushi and Nimmo' (line 12). She then 
resumes her turn, disclosing personal information and relating a 
specific party event. Madhu prefaces this forthcoming tale with us 
din part,! me 'at a party one day' (line 12). Vasant obviously does 
not make an effort to notice this marker_of_extended talk for he 
abruptly intervenes with a new subject, Ushi (line 13). Although it 
may appear Madhu has completed her turn, she has not given up the 
floor but attempts to continue her talk, this time proceeding with 
Vasant's newly initiated discourse topic of their cousin. Moreover, 
it is obvious that Vasant does not interpret the progressive 
development of Madhu's topic "manners and appropriate behavior" but, 
in fact, appears to ignore it. He focuses only on her final ^ 
statement Jise be^hne , uthne, bolne ka sallka naht, yo admT nahT - 
pashu hi. 'one who doesn't know how to act is not human - but an 
animal' (line 11). Interpreting this utterance as a personal 
attack, without warning, Vasant abruptly shifts the topic again to 
pashu! to turn mujhe pashu samajhtl lio? 'Animal! So you think I am 
an animal?' (line 15). 

II. ^ _ - ~ _ « _ 

11. V: ..." mt jab hasta hu, jT khol kar hasti hu or isiliye 

ushTpr nimmo . . . _ _ 

When I laugh, I laugh openly and that's whv Ushi and 
Nimmo . . . ,». _ _ 

12. M: apng phir ushT ?r nimmo kl bat cherT . muJhe hasna bura 

nahl lagta. par samay-kusamay ka bhl dhyan hona cahiye. 
us din part! me ate ^F ushi ne mere kin par cutakl le 
ll 3r nimmo ne raerT akhe band kar ll. koI_samay tha 
us tarah ke hSsi-mazak ka. muJhe hasi-mazak se nafrat 


nahl, badtraizi se nafrat ht. 

Again you bring Ushl and mmmo up . I don't hate 

laughing. But keep in mind the time and the place. 

At a party one day , Ushi covered my ears 

and Nimmo covered my eyes. It wasn't the time for 

such a joke. I don't hate laughing, but I do 

hate bad manners. 

13. V: ushT. .. 

Ushi ... ^ 

14. M: parle sir kT badtmlz hi., madan kT vars-gath ke din 

vo sab aye the. nimmo itnT_cancal_ht, par vo to 

btth gayi ek taraf; ye na^abzadi a b£thT 

mere samne s£ndal saraet, tage pasare or vo uske 

gande s£ndal - merl sifT ke bilkul nazdlk a gaye! 

ap is badtmlijT ko shak se pasand kar?, mt ise hargiz 

bardasht n^f kar saktl.^ .1ise bt.thne, uthne, bolne ka 

sallka nahl, vo adml nahl - pashu hj . 

On the other hand, she has bad manners. They all came 

to my birthday party. Nimmo is usually restless but 

she sat on one side; but the princess 

sat right in front of me with her sandals, 

sprawling and those dirty sandals - they came so close 

to my sari. You may like this bad behavior, but I can't 

approve of it. One who doesn't know how to sit, stand, 

and talk is not human - he/she is an animal . 

15. V: pashu ! to turn mujhe pashu samajhtT ho? tum^admT kT 

sahaj bhavanao ko nirmam varjnao ki beriyo me £se 
bSdh kar rakna cahtl ho... 

Animal I Then you consider me an animal? You want to 
keep a person's natural behavior tied up in chains of 
rules. . . 

To gain a clearer understanding of the linguistic systems of 
topic flow and topic shift, passage III is provided from Ashk 
(1976:75-76). The two participants Madhu (M) and Vasant (V) are 
discussing the subject of cleanliness and filth. On the female 
speaiker's (Madhu) comparison of cleanliness with poetry and art 
(line 16), the male speaker (Vasant) expresses impatience and 
abruptly interrupts her, changing the discourse topic to and 
continuing on a more narrowly defined subject, that of 'hatred for 
filth' (line 17). Although in this passage Madhu interjects with 
side comments (e.g., to phir kure ke dhero par bgthiye 'then go sit 
on a gajtbage pile' (line 20)) and question forms (e.g., ra£_ kab 
kahtT hu? 'When did I say that?' (line 26)), for the most part, she 
remains silent while Vasant is absorbed in his particular discourse 
topic. It is not until Vasant adds his final touch of samjhi ! 
'understand!' (line 23) that the floor is formally turned over to 
Madhu. Here his use of this emphatic question is not a request for 
a response, but an understood signal that he is closing his turn and 
that his partner should take the floor. 



16. M: 





24. M: 


\ mZ kahhT hii l, ap unke svabhav se parlcit nahl, islliye 

ap ko bura laga. svacchata or suruci kl bhavna bhi 

kavya ar kala hT kl tarah... 

Il sayt you felt bad because you didn't understand his 

behavior. Cleanliness and good taste are like poetry 

and art. . . _ ^ 

kyo kavya or kala ko apnT is ghrna me ghasI^tT ho. 

tumhare-£se vatavaran me pal^e hue sab logo kl suruci 

me" ghpna kT bhavna kam karti 

hfc - sharTr se, gandagi se, zindagT se ghpna kT! 

Why do you drag poetry and art into your hatred? 

Underneath the good taste of all people brought up 

like you is a hatred for the body, for filth and for 


(p.up rahti h{^. 

(remains silent)! 

ir mujhe gindagT se ghrna nahX. mujhe sharlr se bKI 

ghrna neihT or mt sac kah du, mujhe gandagi se bhi 

ghriia naht. 

And life isn't hateful to me. I don't hate the body, 

and to tell you the truth I don't hate filth either. 

to phir kure ke dherS par b£^hivel 

rhen go sit on a garbage pile I 

mujhe gandagi se nafrat naht, lekin mt gandagi pasand 

nah? karta -_ba£a sukjma-sa" antar ht. agar 

zindagT ka samna karna hi to roz gandagi 

se do-car Jiona paregi. phir isse nafrat k£.sT?_ 

Jin garlbo ko turn a£ne baramde ke farsh par bhi ^ 

pav na reikhne do, mt unke pas ghanto b6th sakta hu. 

I don't hate filth, but I don't like It'either - 

there's a big difference - if we're going 

to face life then we have to face 

dirt everyday. How can one hate it? I can sit 

for hours with poor people whose feet you won't allow 

on vour veranda. 

(Keval hasti hfc] 
She only laughs. 

3r mt ne £se gande ilako rae zindagT ke laga tar kal varg 
bitaye ht, Jaba-Uimhari suruci kT sanak tumhe guzarne 
teik na de . I sam 1hT ! I 

And I have spent years in dirty areas where you, 

with your good taste wguldn't pass through . lUnder stand \ \ 
par ab to ap garlb nahT. ab to ap gande ilako me nahl 
rahte. garlbl kT mazburl mt samajh saktl hu, lekin 
gandepan ki svabhav merl samajh se dur ki clz hL. 
But now you aren't poor. Now you don't live in dirty 
areas. I can understand if you're poor, but I don't 
understand how filth can^be a part of one's nature, 
to mt svabhav se ganda hu? 

■ Sq I'm naturaUy dirty? 

\mZ kab kahtl hu? | 


Iwhfin did T .-^ av that? (^ 
27. V: tse din raujh par aye ht. 
I have had days .... 

Moreover, by examining Hindi cross-sex conversations such as 
passages I-III further, it is identified that the female's 
participation in the discussion tends to center around joint-sharing 
of experiences, providing support and confidence, offering 
reassurances, and claiming advice and counsel without being hostile, 
dogmatic, or authoritative. Whereas the male speaker tends to act 
as an expert or lecturer, giving advice, expressing facts and rules, 
and explaining reasons and predictions. 

4.0 Other linguistic mechanisms examined which Illustrate that 
female and male speakers effectively use language differently in 
Hindi are the many organizational devices used by the speakers to 
regulate verbal contact and conversation. To maintain conversation, 
bits of talk, vocalizations, and other related linguistic behavior 
are evidence of attention, Interest, and understanding on the 
listener's part. Not being major contributors to the content of the 
conversation, these devices add to the progress and flow of the 
verbal interaction. 

A number of these cross-sex strategies are available to male 
and female Hindi speakers. Some widely used dlscoursal skills of a 
female speaker which coordinate and organize verbal interaction and 
fill in dlscoursal space. Include the following. (Examples are 
squared in passages I-III above and passages IV-VI below.) 1. The 
female speaker employs the strategy of formulaic expressions to 
maintain a steady flow of conversation. Throughout the Hindi 
passages, the following linguistic fillers^are interspersed in the 
f^ale speech: routine responses such as mt kahtT hu 'I say', ml 
puchtl hu 'I ask', and m^ janti jiu 'I know'; fixed expressions such 
as ^ ag kah sakte ht ^You can say this'; and markers of surprise 
such as^nfg. kab kahtl hu? 'When do I say that?' and mt kab inkar 
karti hu 'Do I ever deny It?' 2. A statement uttered by the male 
speaker is partially repeated or briefly rephrased by the female 
speaker.^ For example, in passage VI below, the male speaker Vasant 
states mt ne fcneik utar rakhl ht jr fcnak ke bina turn jantT ho, hamarT 
duniya. . . 'I don't have my glasses and you know without my glasses, 
mj world'... (line 49); the female listener Madhu interrupts (line 
50), affirming her attentiveness with honorific jT and reaffirming 
her linguistic cooperation with the partial repetition of Vasant 's 
final statement 1g kT duniya ! 'your world!'. She then elaborates 
with jane ag kis duniya me rahte h^! 'who knows in what world you 
live'! Similarly, in passage I above, the female speaker Mallika 
seeks clarification of a previous statement uttered by her male 
partner, Kalidasa. Kalidasa claims there Is a big question 


confronting hira, usse kahf bara prashna mei^e samne ht 'I have a 
bigger question than that' (line 2). Mallika responds, rephrasing 
the male spe^er'^ statement and requesting elaboration: or vo 
prashna m£ hu. . .hu na? 'And I am the question, aren't I?' (line 3)- 
3. A sentence begun by the male speaker is completed by the female 
partner. In passage VI below, the male speaker Vasant begins a 
statement in Hindi rat ne apne sabhT patte — 'I had already...', 
(line 55), only to be concluded by Madhu, the female listener, who, 
exhibiting and proffering that she holds the same cultural 
understanding as he, supplies the rest mez gar rakh diye the ! 'laid 
your cards on the table!' (line 56). Not relinguishing the floor, ^ 
she accompanies tj}is thought with a second organizational device mfe 
kab inkar kartT hu 'When did I deny it' (line 56). Similarly, 
earlier in the passage (lines H^^HU) Vasant begins a statement bat 
j[e htki madan ke ta liye chote hfc or.. . 'the problem is that your 
towels are small and...'; and Madhu completes it hazamat ke t llyo ^ 
jfcse h^'made of terry cloth'. 4. To begin a turn the female 
speaker explicitly acknowledges the previous utterance or discourse 
topic spoken by her male partner with the use of acceptance and 
non-acceptance responses to questions and non-requests. Common 
responses include, among others, affirmatives ha 'yes', jT 'yjgs', 
accha 'good', sac 'true', zurur 'to be sure', and denials neihT 'no' 
and bas 'enough'. The female speaker then links her male partner's 
preceding statements with her own by elaborating on his previous 
utterance or on a related topic. This strategy serves to affirm the 
existence of her male listener and his discourse topic. In passage 
I (lines 7-10) above, Mallika verbally expresses her recognition of 
Kalidasa's questions addressed to her. She explicitly provides 
responses of emphatics hS 'yes' (line 8) and nahi 'no' (10) and 
information directly complementing the preceding inquiries of 
Kalldasa. In addition, Madhu emphasizes the subject matter ^^ her 
male partner by using such responses as positive remarks ( ha 'yes', 
accha ) , politeness markers ( caliye 'go', ap 'you') and assenting 
deference elements ( jT ' hon . ' ) , to name a few. 


28. "tum ab sukhl ho?"_ uskajsvar bahut dhTma_tha. 

29. ''hain dono pahle bhl sukhl the" - usne kaha. 

30. 'pgU..lekin ab tum sukhl ho?" 

31. ' (turn 1antl ho U-ye ham dono ke liye thik ne tumse 
pahle bhl kah tha." 

"Are you happy now?" her voice was soft. 

"We w ere both happy even before," he said. 

lYes^ . .but are you happy now?" 

■ [you know [...this was right for both of us... I told you 

that before . " 

(Verma, Nlrmal: an tar) 


32. V: Ikaho bhalj kya hal-cal ht? ye surat kisT ronT banS 
rakhl hfc. ]T ku ch kharab htkya? 
[Tell me brother ! how are you? Why the long face? 


zukam hfc mujhe tin- car 

Do you feel ok? 
33 • M: sukhi jira par raha h£. 
din se. 
It's the dryness of winter. I've had a cold for 3-'' 

. [1a,Y.1. = = , 

34. V: I mfc ne tumse kitnT bar kaha hu ki apnT s ehat ka dhyan 

rakl;;S karo. sehat -sehat- sehat! .. j areTl ve kayapala^ 

k6si! ye palag draig-rum me kfcse a gaya. dt tre or 

pvale . — '_ 

|How many times have I told you| t o ta ke care of your 

health? Health-health-health! .. jQhJ These changes? 

How did the bed get in the drawing room? 

And a tray and cups.... _ 

35- M: mfc. ne palag idhar hT bichi diya ht, ki ap or ap ke mitro 

ko zar¥ bhT kast na ho. maze se lihaf le kar bfctiye. 

tellfon apke sirhane rahega. 

I had the bed moved in here so that you and your friends 

would not have any problems. So that anyone could sit 

comfortably under the quilt. The phone will 

be at the head oC t he bed . . , 

V: vah! mfc kaht a hu. jtum tol .J tum to^ .. behad acchi ho! 

[Oh! I I say! I You.I voul are wonderful! _ 
M: mt khud apnT saheliyo ke sath isT lihaf me* b€thi rahl 


I've been sitting under the quilt with my friends. 
V : sac ! 

M: lacchal. ap jakar hat-muh dho lijiye. mt day tiyar kartT 


[Qk|. go wash your hands and face. I'll get tea. 
V: mb kahta hu ki tun kitnT...tum kitnl turn kitnT acchT 


I sav. you're so... yo u 're so... you 're so good! 
M: laccha. accha. caliyel pjihale hat-muh dho kar kapre 





Ik. ok. got First wash up then change yur clothes. 
(Ashk 1976:91-93) 


42. M: 

43. V: 

44. M: 

iL kahtT hul ap. 

sav| you. . . _ ^ _ 

Oh! I ve kambaxt_t3liye! mujhe dhyan hT nahT rahta. 
bat ye hf. (hasta he) ki madan ke tJlive chote he or ... 
| 0h!| Those stupid towels! I didn't remember (laughs) 

the problem Is y£ur towels^arg small gr^d — 

hazamat ke tJliyo - It-se h£ .. |jl!| |zari akh khol kar l 
[dekhiyej hazamat ke tJliye kitne ragTn h£., 
bTsiyo to dhariya papT hul ht 

unjne or madan ke kitne s ide 3 r . . . 

...made of terry cloth. JYesJ I Please lookl - how 
colorful your shaving towels are, there are many stripes 
on them while mine are plain and... 


45. V: 

46. M: 

47. V: 

48. M: 

49. V: 

50. M: 











lekln royedar to... 

But^they are t erry cloth... ^ _ 

dono ht, ! 1 ll^l akhe ban d karke IdmT dono ka fark bati 

sakta ht! Imtkahrr huj . . 

Both are! lYesil Even w ith eyes closed one can tell 

the difference. |l sayj. . . 

asal me mera dhyan dii^rT ar tha. | laol mujhe hasamat_ 

ka tDliya de do. kaha ht, mujhe dikhayT hi nahl diya. 

My mind was somewhere else. Ok, | give me| my towel. 

I couldn't find it. 

ye to taga hi, samne, phir bhT. . . 

U's hanging right in front of you, if only..._ 

mV ne <-nak utar rakhl ht jr cnak ke bina turn jantFho, 
IhamarT duniva.l . . 

I don't h ave my glasses and you know without my glasses, 
[my world J. . 
dr.! ap kl duniya! lane ap kis duniya me rahte hfc ! 

ab to fcnak nahi, tnak ho to kon-sa apko kuch 

dikh ayT deta ht! 

lYesJ your world! Who knows in what world you live! 

Now you don't have your glasses on, but even when you 

do, you don't see anything anyway! 

ye phir tumne rauh phula liya, naraz ho gayr ho? 

You' re pouting again, are you mad at me? 
[nahilmt naraz nam . 

No, I I'm not mad. . „ v 

turahara khyal hfc, ki mt itna murkh hu, Jo ye bhT nahT 

pahcan sakta? 

Do you think that I am that stupid that I can't 

56. M: 

When did I say that?| ^ 

mfc ne tumse kitnT bSr kahd- ht l ki apne bhavo ko chipa 
lena tumhSre bas kT bat nahf. tumharT nafrat, 
tumhara krodh, tumharT sarT bhavnae, tumhare 
cehare par jhalak ^atT ht.^tumhe merT^ 
adate burT lagtT h^ par mt_ne J^umhe adhere me nahl 
rakha. apne bare me, apnT adato ke bare me, sab kuch 
bata diya tha. mt ne apne sabhT patte . . . 
Ihow many times have I told you l that you can't hide your 
feelings, even if you try. Your hatred, your anger, 
and in fact all your feelings show in your face. 
You don't like my habits, but I never kept you 
in the dark. I told you about myself, my habits, 
and in fact, about ever ything. I had already,... 

mez par rakh diye the! I mt kab inkSr kartT hu 1 . | 

laid your cards on the table ! I Did I every deny it? I 
(Ashk 1976:70-71) 

This is not to say that the male Hindi speakers do not hold up 


their ends of conversation in such a manner, but that the dlscoursal 
strategies which they empower to coordinate turns and organize the 
flow of speech are different and less frequent. (Examples are 
squared and refer to passages I-VI above). 1. Where the female 
speaker uses first person conventions such as mt kahtT hu 'I say' 
and mt. puchtT hu 'I ask', the male speaker addresses to gain and 
hold his listener's attention with the direct measure of second 
person expressions such as turn 'you', turn jantT ho 'you know', and 
samjhi 'you understand'. 2. Similarly, the male speaker's use of 
second and third person pronouns turn and ham, 'you' and 'we', 
respectively, is a device to acknowledge and include in the on-going 
discussion the existence of the female listener. In passage VI, the 
male speaker Vasant utters turn JgntT ho, hamarT duniya " (line 43). 
This is literally translated as 'you know, our world' but Is 
interpreted to mean "my world". S* The male discussant uses 
spesiker selection elements to explicitly label and to attract, 
catch, and hold the attention of his female listener. Such 
linguistic attention getters include, among other constructions: 
kaho bhir 'tell me brother' and turn jantT ho 'you know'; imperative 
markers: lao 'bring' and soco 'think'; address terras: the listener's 
personal given name (e.g., Mallika ); address and reference forms: 
bhir 'brother', ma^am , and turn 'you'; and evocative markers of 
surprise: oji, are 'hey', uh 'uh', and yah. Although the female 
speaker similarly uses the above linguistic markers, i.e., 
imperative forms such as dekho 'see' and suno 'listen', the male 
forms are used as attention getting devices to solicit and secure 
the attention of his listener, while the female forms function 
similar to boundary markers, fillers, or accompaniments which 
regulate and maintain conversation. U. Where, on the one hand, the 
female speaker tends to begin sentences with verbal acknowledgments 
the male speaker tends to ignore the preceding comment and pursue 
his own topic to effectively hold the floor. As shown in passage 
III above, the male speaker (Vasant) does not acknowledge the female 
speaker's (Madhu) side comments but continues on as if no 
interruption or remark was made. Even vocalizations by his female 
partner such a§ laughter, snickers, and the use of fillers such as 
mg kab kahtT jiu 'When did I say that?' do not deter his continuation 
and flow of discourse. To provide further evidence, after each of 
Madhu 's comments, as if ignoring her verbal presence, Vasant 
initiates turns with the conjunction aur 'and', as if making a claim 
his turn is still in progress. It is found that the male speaker 
begins his turn in this manner five times greater than the female 
speeiker, all incidences of which are after responses such as these 
conversational maintaining devices. 5. The male speaker tends to 
use interrogative forms which appear as imperatives, articulating an 
immediate concern by the speaker and requiring prompt attention by 
his listener. The use of these forms draws attention to the speaker 
and exhibits the convention of expert knowledge. Such samples from 
the male speech in Hindi, among others, include m£ ne tumse kah tha 
'Didn't I tell you...' (IV, 31) and mt ne tumse kitnl bar kah ht 'How 
many times did I tell you...' (VI,55T. 


For the female Hindi speaker, such organizational devices are 
signals of continued attention that she is listening to and 
following closely what is being said by her male partner. Not only 
do these markers provide the means for her to actively participate 
in the conversation and to keep the general coordination of speech 
going, but they also afford a sense of courtesy that her male 
partner should resume and produce a fuller talk. She is claiming 
attention as well as calling for elaboration and further development 
of the discussion. By no means, however, are these mechanisms 
attempts to take over the turn. For the male speaker, 
organizational devices are in the form of listener Inclusive 
elements, speaker selection items, expert knowledge expressions, and 
attention getting markers. They function as a means to ensure that 
his female partner is listening, following, and attending to his 
remarks. For both male and female Hindi speakers, then, 
organizational devices are simply reciprocal discoursal measures. 
They are symbols of strength which act as a means to ensure an 
effective cross-sex conversation. 

5.0 Conclusion. Since male and female speakers of Hindi have 
different experiences and social roles and needs in India, it is 
only expected that the sexes develop different strategies and skills 
of speech to operate in the society. In particular, I have 
suggested that when the female and male Hindi speakers Interact in 
cross-sex conversations they use verbal strategies differently. 
Maltz and Borker (1982) argue that because male and female speakers 
come from different sociolinguistic subcultures, these language 
users hold different models for friendly cross-sex conversation. 
This view possibly explains why female Hindi speakers are the 
maintainers and sustainers of conversation, whereas for the male 
Hindi speakers, speech functions more as a means of relating events, 
conveying messages, and gaining information and attention. 

The examination of cross-sex interactions in Hindi reveals that 
a differential use of discoursal strategies is held by male and 
female Hindi speakers. First, the speakers have different rates of 
successful initiation of discourse topic, both with a question and 
with a statement. For the female speakers, questions and statements 
function as conversational maintenance devices; for the male 
speakers, questions are interpreted as requests for Information, and 
statements suffice as forms to initiate successful conversational 
topic. Second, the female speakers develop and shift conversational 
topic slowly and gradually; the male speakers narrowly define topic 
and shift abruptly. And, third, to regulate and coordinate 
cross-sex verbal interaction, a number of linguistic devices are 
employed by the female and male participants. As an expression of 
attentiveness and understanding, sentences begun by the male 
speakers are completed, partially repeated, restated, clarified, and 
verbally emphasized and acknowledged by the female partners. As a 


direct means to gain attention and a response, the male speakers 
ignore their partners' comments and efforts for a smooth stream of 
discourse, express expert knowledge, indicate the existence of their 
female listeners, and receive verbal deference, acknowledgements, 
and attention. 

By no means do I propose that all cross-sex conversations are 
ideal speech situations. We see the frustration and exasperation 
speakers such as Madhu experience when Vasant makes accusations 
about her beliefs, or when questions are repeated a number of times 
by Mallika to gain a response from her male partner. That each 
gender uses and interprets discoursal strategies in Hindi in varied 
ways, suggests that the power strategies of each sex do come into 
conflict and can cause miscommunication (Tannen 1982). What is 
important, however, is that these discoursal strategies are symbols 
of power that do not necessarily require the domination of one 
language user over another. The fact that a particular linguistic 
skill is used as a resource of interaction and plays a different 
role for each sex does not mean that the speaker who makes use of it 
more frequently is socially powerful or powerless or 
conversationally secure or Insecure, but that this verbal strategy 
is an expression of strength in itself and is motivated and 
interpreted differently by each sex for the mutual goal of 
effectively communicating in cross-sex discourse. 


ASHK, Upendranath. 1976. T^liye. In Mere prlya ekanki, pp. 67-95. 

Allahabad: Rachna Prsikashan. 
COULTHARD, Malcolm. 1977. An introduction to discourse analysis. 

London: Longman. 
ELSHTAIN, Jean Bethke. 1982. Feminist discourse and its discontents: 

language, power and meaning. Signs: Journal of Women in Culture and 

Society 7:3-603-621. 
FISHMAN, Pamela M. 1980. Conversational insecurity. In Giles,, 

eds.: pp. 127-132. 
FISHMAN, Pamela M. 1983. Interaction: the work women do. In Thorne,, eds.: pp. 89-101. 
GILES W., Peter Robinson and Philip M. Smith, eds. 1980. Language: 

social psychological perspectives. New York: Pergamon Press. 
HARTSOCK, Nancy. 198I. Political change: two perspectives on power. 

In Quest Staff and Book Committee, eds.: Building Feminist Theory, 

PP' 3-19. New York: Longman. 
KEENAN, Elinor Ochs and Bambi Schleffelin. 1976. Topic as a discourse 

notion: a study of topic in the conversations of children and 

adults. In Charles Li, ed.: Subject and topic, pp. 335-384. New 

York: Academic Press. 


LEET-PELLEGRINI , H. M. 1980. Conversational dominance as a function of 

gender and expertise. In Giles,, eds.: pp. QY-ICI. 
MALTZ, D. and R. Borker. 1982. A cultural approach to raale-female 

misconununication. In John J. Gumperz, ed.: Language and social 

identity pp. 196-214. Cambridge: Cambridge University Press. 
RAKESH, Mohan. 1958. "Asharh ka ek din. Delhi: Rajpal and Sons. 
SACHS, Harvey, Emanuel Schegloff and Gail Jefferson. 1974. A simplest 

systematics for the organization of turn-taking for conversation. 

Language 50.696-735. 
SATTEL, Jack W. 1983. Men, inexpressiveness and power. In Thorne,, eds.: pp. 118-124. 
SOSKIN, William F. and Vera P. John. I963. The study of spontaneous 

talk. In Roger Barker, ed.: The stream of behavior. New York: 

Appleton-Century-Crof ts . 
TANNEN, Deborah. 1982. Ethnic style in male-female conversation. In 

John J. Gumperz, ed.: Language and social identity, pp. 217-231. 

Cambridge: Cambridge University Press. 
THORNE, Barrie, Cheris Kramarae, Nancy Henley, eds. 1983- Language, 

gender and sex. Rowley, MA: Newbury House. 
VERMA, Nirmal. 1964. Antar. In Jalti jhari, pp. 142-155. Delhi: 

Rajkamal Prakashan. 
WEST, Candace and Don H. Zimmerman. 1983- Small insults: a study of 

interruptions in cross-sex conversations between unacquainted 

persons. In Thorne,, eds.: pp. 102-117. 
ZIMMERMAN Don H. and Candace West. 1975. Sex roles, interruptions, and 

silences in conversation. In Barrie Thorne and Nancy Henley, eds.. 

Language and sex: difference and dominance, pp. 105-129. Rowley, 

MA: Newbury House. 

The following special issue is in preparation: 

Linguistic Studies in Memory of 

Theodore M. Light ner 
Editors: Michael J. Kenstowicz 
Charles W. Kisseberth 


The following issues are available: 
Spring 1977 Vol. 7, No. 1, Papers in General Linguistics 

Fall 1977 Vol. 7, No. 2, Studies in East Asian Linguistics 

Editors: Chin-chuan Cheng and 
Chin-W. Kim 

Spring 1978 Vol. 8, No. 1, Papers in General Linguistics 

Fall 1978 Vol. 8, No. 2, Linguistics in the Seventies: 

Directions and Prospects 
Editor: Braj B. Kachru 

Spring 1979 Vol. 9, No. 1, Papers in General Linguistics 

Fall 1979 Vol. 9, No. 2, Relational Grammar and Semantics 

Editor: Jerry L. Morgan 

Spring 1980 Vol. 10, No. \, Papers in General Linguistics 

Fall 1980 Vo. 10, No, 2, Studies in Arabic Linguistics 

Editor: Michael J. Kenstowicz 

Spring 1981 Vol. 1 1, No. 1, Papers in General Linguistics 

Fall 1981 Vol. 1 1 , No. 2, Dimensions of South Asian 

Editor: Yamuna Kachru 

Spring 1982 Vol. 12, No. 1, Papers in General Linguistics 

Fall 1982 Vol. 12, No. 2, Papers on Diachronic Syntax: 

Six Case Studies 
Editor: Hans Henrich Hock 

Spring 1983 Vol. 13, No. \ , Papers in General Linguistics 

Fall 1983 Vol. 13, No. 2, Studies in Language Variation: 

Nonwestern Case Studies 
Editor: Braj B. Kachru 

Spring 1984 Vol. 14, No. 1, Papers in General Linguistics 

Fall 1984 Vol. 14, No. 2, Language in African Culture and Society 

Editor: Eyamba G. Bokamba 

Spring 1985 Vol. 15, No. 1, Papers in General Linguistics 

Orders should be sent to: 

SLS Subscriptions, Department of Linguistics 

4088 Foreign Languages Building 

707 S. Mathews 

University of Illinois 

Urbana, IlHnois 61801 

Studies in 
The Linguistic 


^ 1 n 



CATHERINE V. CHVANY Ergative and argative (nee ergatiye 

too) 1 

HERBERT S. COATS Palatalization in Russian 3 

ALICE DAVISON Case and control in Hindi-Urdu 9 

JOSEPH F. FOSTER Primitiveness, naturalness, and cultural 

fit 25 

FRANK Y. GLADNEY On glides following vocalic verbs in 

Russian 39 

MORRIS HALLE Remarks on the scientific revolution in 

linguistics 1926-1929 61 

MICHAEL KENSTOWICZ The phonology and syntax of 

wh-expressions in Tangale 79 

CHIN-W. KIM Phonology on the "C-string"? 93 

F.K. LEHMAN with NAMTIF PINGKARAWAT Missing nominals, 
non-specificity and related matters, with especial reference to 
Thai and Burmese 101 

WINFRED P. LEHMANN The persistence of pattern in language 123 

ZHIJI LU and CHIN-CHUAN CHENG Chinese dialect affinity 

based on syllable initials 1 27 

HORACE G. LUNT On the progressive palatalization of early 

Slavic: synchrony versus history 149 


EDWARD J. VAJDA Derived imperfeclives in Slavic: 

a study in derivational morphology 171 

CARLOTA S. SMITH Sentence topic in texts 187 

ARNOLD M. ZWICKY The case against plain vanilla syntax 205 

Department of Linguistics 
University of Illinois 





EDITORS: Charles W. Kisseberth, Braj B. Kachru, Jerry L. Morgan 

REVIEW EDITORS: Chin-W. Kim and Ladislav Zgusta 

EDITORIAL BOARD: Eyamba G. Bokamba, Chin-chuan Cheng, Peter 
Cole, Alice Davison, Georgia M. Green, Hans Henrich Hock, Yamuna 
Kachru, Henry Kahane, Michael J. Kenstowicz and Howard Maclay. 

AIM: SLS is intended as a forum for the presentation of the latest original 
research by the faculty and especially students of the Department of 
Linguistics, University of Illinois, Urbana-Champaign. Especially invited 
papers by scholars not associated with the University of Illinois will also be in- 

SPECIAL ISSUES: Since its inception SLS has devoted one issue each year to 
restricted, specialized topics. A complete list of such special issues is given on 
the back cover. The following special issue is under preparation: Illinois 
Studies in Korean Linguistics, edited by Chin-W. Kim. 

BOOKS FOR REVIEW: Review copies of books (in duplicate) may be sent to 
the Review Editors, Studies in the Linguistic Sciences, Department of 
Linguistics, University of Illinois, Urbana, Illinois, 61801. 

SUBSCRIPTION: There will be two issues during the academic year. Requests 
for subscriptions should be addressed to SLS Subscriptions, Department of 
Linguistics, 4088 Foreign Languages Building, 707 S. Mathews, University of 
Illinois, Urbana, Illinois 61801. 

Price: $5.00 (per issue) 
Price increase to $6.00 
beginning with Vol. 16:1. 

The person charging this material is re- 
sponsible for its return to the library from 
which it was withdrawn on or before the 
Latest Date stamped below. 

Theft, mutilation, and underlining of books are reasons 
for disciplinary action and may result in dismissal from 
the University. 

To renew call Telephone Center, 333-8400 


m 13 1981 


APR 07 
AUG 1 





L161— O-1096 




Catherine V. Chvany: Ergative and argative (nee ergative too) ... 1 

Herbert S. Coats: Palatalization in Russian 3 

Alice Davison: Case and control in Hindi-Urdu 9 

Joseph F. Foster: Primiti veness, naturalness, and cultural fit. . . 25 

Frank Y. Gladney: On glides following vocalic verbs in Russian. . . 39 

Morris Halle: Remarks on the scientific revolution in linguistics 

1926-1929 61 

Michael Kenstowicz: The phonology and syntax of wh-expressions in 

Tangale 79 

Chin-W. Kim: Phonology on the "C-string"? 93 

F.K. Lehman with Namtip Pingkarawat: Missing nominals, non- 
specificity and related matters, with especial reference to 

Thai and Burmese 101 

Winfred P. Lehmann: The persistence of pattern in language 123 

Zhiji Lu and Chin-Chuan Cheng: Chinese dialect affinity based on 

syllable initials 127 

Horace G. Lunt: On the progressive palatalization of early Slavic: 

synchrony versus history 149 

Lew R. Micklesen, Sally R. Pitluck, and Edward J. Vajda: Derived 

imperfectives in Slavic: a study in derivational morphology. . 171 

Carlota S. Smith: Sentence topic in texts 187 

Arnold M. Zwicky: The case against plain vanilla syntax 205 


Theodore M. Lightner was born on September 5, 1934 and died 
suddenly and unexpectedly in March 1984. He took the B.S. degree 
at Duke University in 1958 and the Ph.D. in Linguistics at MIT in 
1965. He held teaching positions at the University of Illinois 
(1965-1969), the University of Texas (1969-1973), and several 
European universities including Aix-en-Provence and Trondheim. 
He was the author of over forty scholarly publications, including 
two books. Lightner had wide-ranging interests in linguistics 
including phonology, morphology, syntax, semantics, mathematical 
linguistics, Slavic, and Indo-European. His most significant 
contributions lay in the areas of phonology and morphology. 

Throughout his career Lightner was concerned with one basic 
problem: how to characterize the relation between grammatical 
units (morphemes) with the same or similar meaning and differing 
phonetic realizations. His answer was that if there was reason 
to believe that the relation was systematic, then it should be 
stated in terms of derivation from a unique underlying form. He 
saw no reason to believe that there are any inherent constraints 
on the degree to which the different phonetic alternants may 
depart from the underlying representation. This position, which 
he consistently maintained throughout his career as a linguist, 
had a very significant impact on the development of generative 
phonology. In their early work, Chomsky and Halle showed that 
concern for economy and generalization required analyses that 
violated structuralist preconceptions on proper phonological 
representations (e.g. Chomsky's derivation of (dialectal) English 
can't [kit] from /kant/ or Halle's celebrated argument against 
the phoneme on the basis of Russian voicing assimilation). While 
these analyses violated taxonomic phonemic canons, they were well 
within the spirit of Sapir or the descriptive Bloomfield. 
Lightner forced generative phonologists to see that the same 
concern for economy and generalization also motivates a level at 
which "deeper" alternations such as the vowel shift in English 
sane-sanity derive from a unique underlying representation — an 
alternation that even Sapir would have treated by the listing of 
allomorphs. The most important effect of this move was to 
considerably expand the data base upon which phonological 
analysis in terms of underlying representations and ordered rules 
could be based, thereby making possible most of the important 
innovations of early generative phonology (e.g. the notational 
conventions, the cycle, rule ordering, etc.). It is likely that 
the field would have evolved to this point sooner or later 
anyway, but it is fortunate that it did so so early. Lightner 
must be given a good deal of the credit for this. 

Lightner's early work was not on English but Russian. Here he 
showed how many deep alternations traditionally treated by the 
listing of morpheme alternants could be brought under the rubric 
of unique underlying representations. His analyses exhibited 
considerable insight and imagination, as in the famous 
description of the yers. Lightner's work on Russian culminated 
in his first book Problems in the Theory of Phonology (1972). 

This treatise has been the single most important contr ibui tion to 
Slavic linguistics by a generative phonologist; all subsequent 
generative phonological work on Slavic begins by agreeing or 
disagreeing with the analyses presented in this book. 

Ted Lightner had a combative personality, both personally and 
professionally. His style was to take a strong, often extreme, 
position and dare you to prove him wrong. One of his favorite 
locutions was "I bet you ten dollars that...". Encounters with 
him thus often turned into intellectual arm-wrestling contests. 
(I am told that he was a varsity wrestler at Duke.) Of course, 
this kind of stance is often the appropriate one to take when one 
is trying to develop a radically different approach. It helps to 
clarify the issues and to attract supporters, especially among 
the young. During the late ' 60 ' s when he was at the University of 
Illinois Lightner was one of the most active and heavily 
recruited generative linguists and attracted a good many students 
to the study of linguistics, especially phonology. However, 
during the '70's the field of phonology and generative grammar as 
a whole changed to take up the question of constraints on rules 
and representations. This was a line of research to which 
Lightner did not contribute, feeling that not enough descriptive 
work had been done to be able to address the question of 
constraints satisfactorily. Rather, he continued to develop, 
essentially in isolation, his concern with characterizing all 
phonological relationships in terms of unique underlying 
representations. This line of thought naturally led him to 
Indo-European and culminated in his final book Introduction to 
English Derivational Morphology (1983). This strange book is 
probably best known for proposing that the roots in such English 
words as father and paternal derive synchronically from the same 
underlying representation. Among other things this move has the 
bizarre consequences that the underlying phonemic inventory of 
English contains laryngeals, and that Grimm's Law is a rule of 
contemporary English phonology. Lightner continued to challenge 
what he saw as the phonological establishment to prove him wrong 
in this and to propose explicit criteria to separate synchrony 
from diachrony. The majority of phonologists would undoubtedly 
respond by saying that this line of research is not as likely as 
others to reveal significant insights into the phonological 
faculty and hence is not a particularly worthwhile question to be 
asking, at least at this point in the development of the field. 
Lightner ended his twenty year career as a linguist in a 
relatively marginal and isolated position. Nevertheless he was a 
central figure in the beginning of Generative Grammar and it is 
the Ted Lightner of these more happy days whom the majority of 
contributors to this volume knew and wish to pay tribute. 

Michael Kenstowicz 

Editor's note: Given the special nature of this volume, the 
papers did not undergo the refereeing process that contributions 
to SLS normally receive. The papers in this issue represent each 
individual author's way of remembering Theodore M. Lightner. 


1962. On ponjat' and obrazovat' type verbs in RUssian. Quarterly Progress 
Report (QPR) 67.177-180. Cambridge, MA: MIT Research Laboratory 
of Electronics. 

1963a. Note on the formulation of phonological rules. QPR 68.187-189. 

1963b. Vowel harmony in Classical (Literary) Mongolian. QPR 68.189-190. 

1963c. Remarksonthe morphophonemic component of Russian. QPR 69.193-199. 

1963d. Nasal diphthongs in Russian. QPR 70.293-297. 

1963e. The shift of s toxin Old Church Slavonic verb forms. QPR 70.298-300. 

1963f. Preliminary remarks on the morphophonemic component of Polish. QPR 

1963g. Reduction of long i in Russian imperative, infinitive, and two singu- 
lar verb forms. QPR 71.235-236. 

1963h. sr/zr clusters in Old Church Slavonic. QPR 71.236-237. 

1965a. Note on the motivation for using transformation rules in phonology. 
QPR 76 271-274. 

1965b. cikliceskix pravilax v russkom sprjazenii. Voprosy jazykoznani ja 

1965c. Segmental Phonology of Contemporary Standard Russian (unpublished 
doctoral dissertation, MIT). 

1965d. On the description of vowel and consonant harmony. Word 21:2.244- 

1966a. Ob al'ternacii e : o v sovremennom ruskom literaturnom jazyke. 
Voprosy jazykoznani ja 15:5.64-80. 

1966b. On descriptions of common Slavic phonology. Slavic Review 24.679-686. 

1966c. On the phonology of Old Church Slavonic conjugation. International 
Journal of Slavic Linguistics and Poetics X.1-29. 

1967a. Sur I'emploi de regies mineures dans la phonologie du russe. Lan- 
guages 8.67-72. 

1967b. On the phonology of Russian conjugation. Linguistics 35.35-55. 

1968a. An analysis of akan'e and ikan'e in Modern Russian using the notion 
of markedness. Studies Presented to Professor Roman Jakobson 
by his Students, pp. 188-200. Cambridge, MA: Slavica Publishers. 

1968b. On the use of minor rules in Russian phonology. Journal of Linguistics 4. 

1969. On the alternation e : o in Modern Russian. Linguistics 54.44-69. 

1970a. Why and how does vowel nasalization take place? Papers in Lin- 
guistics 2:2.179-226. 

1970b. On regular inflectional endings in English. Papers in Linguistics 

1971a. On Swadesh & Voegelin's Problem in Phonological Alternation. UAL 37: 

1971b. Generative phonology. In William Orr Dingwell, ed.: A survey of 
linguistic science, pp. 498-556. College Park, MD. 

1971c. A problem in coexistent phonological systems. Linguistic Inquiry 2: 

1972a. Problems in the theory of phonology: Russian phonology & Turkish 
phonology. Edmonton, Alberta: Linguistic Research, Inc. 

1972b. Some remarks on exceptions and coexistent systems in phonology. In 
D. Worth, ed.: The Slavic word, pp. 426-442. The Hague: Mouton. 

1973a. On the formulation of Grassman's Law in Greek. In S. Anderson & 
P. Kiparsky, eds.: A festschrift for Morris Halle, pp. 128-130. 
NY: Holt. 

1973b. Against morpheme structure conditions and other things. In M. 

Kenstowicz & C. Kisseberth, eds.: Issues in phonological theory, 
pp. 53-60. The Hague: Mouton. 

1973c. Paired and opposing tendencies in linguistic change: five forms of 

the word for 'rouble' in Modern Russian. In B. Kachru, ed.: Issues 
in linguistics: papers in honor of Henry and Ren4e Kahane, pp. 548- 
553. Champaign, IL: University of Illinois Press. 

1973d. Remarks on universals in phonology. In M. Gross, et al. eds.: The 

formal analysis of natural language, pp. 13-49. The Hague: Mouton. 

1974a. A problem in the analysis of some vowel-zero alternations in Modern 
Russian. In M. Flier, ed.: Slavic Forum, pp. 78-87. The Hague: 

1974b. Preliminary remarks on derivational morphology of French. In C. 
Rohrer & N. Ruwet, eds.: Actes du collogue franco-al lemand de 
grammaire transformational 1 , pp. 142-164. Tubingen: Max Niemeyer. 

1975a. The role of derivational morphology in generative grammar. Language 

1975b. Transitive softening in Russian conjugation (with Herbert C. Coats), 
Language 51.338-341. 

1976a. A note on McCawley's review of SPE. UAL 42.79-82. 

1976b. Review of Goals of linguistic theory, ed. by S. Peters. Language 52. 

1978a. Review of Foundation of theoretical phonology by J. Foley. Lin- 
guisticae Investigationes 2:1.219-230. 

1978b. On a model for language. In E. Scatton, ed.: Studies in honor 
of Horace Lunt, pp. 234-239. 

1978c. Review of Noms d'agent et noms d'action en indo-europeen, by Emile 
Benveniste. Linguisticae Investigationes 2:2.429-452. 

1979a. A problem with description of declension in Homer. Studies in Language 

1979b. On the English etymology of star. Linguisticae Investigationes 3:2. 

1979c. Review of Syntactic typology, ed. by W. Lehman. Linguisticae 
Investigationes 3:1, 187-188. 

1983. Introduction to English derivational morphology. Amsterdam: John 

March, 1921 


As the gentle jester of his monsterhood , 
I tell the saga of my orphanhood.. . 

Behind a prince, his clan; behind an angel, hosts; 
Behind each man, a thousand others at their posts, 

A living wall to fall against if faltering, and know 
That there are thousands more to carry on the work. 

A soldier's proud of his regiment, a devil of his swarm; 
Behind the thief's a gang, but back of the jester — his hump. 

So, weary at last of cleaving to the ken of pointing 
Fingers and the summons to do battle, to the continue 

Of the idiot's hiss and the philistine's hee haw, 
— Alone among all, for all, counter to all — 

I stand and send, at take-off turned to stone. 
This rousing call into the heavens' void. 

And by the fire in my breast this pledge is sworn: 
Someday a Charlemagne will hear you, Horn! 

Translation by 
Catherine V. Chvany 
September, 1985 


Studies in the Linguistic Sciences 
Volume 15, Number 2, Fall 1985 


Catherine V. Chvany 

Massachusetts Institute of Technology 

The terra ergatlve has developed a second meaning in the recent 
literature, one that is antonymous with the traditional one. 
I propose keeping the original term in its root meaning, and 
coining for the second meaning the term argatlve (cf . Argon, 
the inactive gas), prefixing the negative _a- to the root these 
terms share with work , energy , erg . 

As Ted Lightner liked to point out, the words energy and work share 
the same root, which is also found in erg , the physicists' unit of work 
(from Greek ergon 'work'), and in ergatlve (from Greek ergates 'doer, 
actant') a term whose root meaning suggests agency. In so-called ergatlve 
languages, the ergatlve case Inflection marks the (most often agentive) 
subject of a transitive verb, while the direct object appears in the 
absolutive (or nominative) case, as does the subject of an intransitive 
verb. For instance, in the sentence pair, Ivan opened the window / The win- 
dow opened, Ivan would be in the ergatlve case, while the gloss of window 
would appear in both sentences In the same absolutive or nominative case. 
Until Keyser and Keeper's 1984 article, some of us Slavists found it 
convenient to refer to such window-arguments as "the absolutive set." 
In Relational Grammar the roughly corresponding term was "unaccusative . " 
The terms "absolutive" and "unaccusative" tacitly included in their 
reference the role of window in The window washes easily . Since Keyser and 
Keeper's evidence for crucial distinctions between window in the first pair 
(which they term "ergatlve pairs"), and window in the latter sentence type 
(which they term "middles"), the traditional terms "absolutive" and 
"ergatlve" have become hopelessly ambiguous. 

The term ergative has undergone an unusual historical change ever 
since Burzio 1981; this change has spread, so that many linguists in the 
United States now routinely use ergative to refer to non- agentive "absolu- 
tive set" arguments. Pesetsky 1982, for Instance, finds interesting 
generalizations that apply in Russian to the absolutive set, which he calls 
"the ergatlve set." Keyser and Roeper' s discussion of ergative pairs 
assigns the term ergatlve to the Inactive rather than the active argument, 
with a note pointing out that in other frameworks their "ergatives" are 
called "absolutive" or "unaccusative." That is, the new sense of ergative 
(let's call it ergative-2) is the antonym of the term in its traditional 
sense (or ergative - 1 ) . 

So far, local communication problems due to the ambiguity have not 
been too serious; one adjusts for metadlalect variations, resorting to 
modifiers, such as "ergatlve in the sense of Burzio" or, as is now more 
likely, "ergatlve in the sense of Keyser and Roeper." But the rise of 
ergatlve-2 does hamper International and interdisciplinary communication. 
Though the results of the research on "ergatIve-2" and "middle" should 
stimulate parallel investigations by linguists all over the world, such 

ambiguities pose obstacles (compounded for those colleagues abroad who read 
English with difficulty). Citation Is awkward If one has to redefine an 
established term; Americans working on, e.g., Slavic should not have to 
explain to Slavs that these "ergatlves" have a precisely opposite role from 
the one they normally associate with the ergatlve case. 

A more transparent term would greatly facilitate the sharing of the 
important findings about "ergatlve-2s" and "middles." For the inactive 
member of Keyser and Keeper's "ergatlve pairs" I propose ARGATIVE, with the 
negative a- attached to the root KG (on the analogy of Argon , the name of 
the Inactive gas.) K&R's "ergatlve pairs" would then be more perspicuously 
renamed "ergative-argative pairs." For native speakers of "ergative-2" 
dialect, the slight phonological change should cause only minimal discom- 
fort. For others it would eliminate an irritant while making the new work 
on grammatical relations more accessible to linguists in other countries, 
including those who study or speak "ergatlve (1)" languages. The term 
absolutive might remain in its traditional sense of "antonym of ergatlve ," 
while NPs like the subjects of "middles" might be called something like 
"argatives with inexpressed generic agent," now that Keyser and Roeper have 
opened our eyes to this distinction among absolutives. 

Massachusetts Institute of Technology 


Slavic dictionaries of linguistic terms associate the term ergatlv 
with agency. Axmanova says it is a case for marking "the source of the 
action." Rozental' and Telenkova's entry for ergatlvnyj describes ergatlve 
sentence of certain (e.g., Caucasian) languages, characterized by "a 
special case for the active doer (the ergatlve, or active, case)." 
The antonymous, complementary case (of our "absolutive set," or 
"ergatlve-2") is termed "a direct case (the passive case of the non-doer, 
having the nominative case form without any case ending). 

The older dictionary of Marouzeau defines ergatif and its German, 
English and Italian cognates as "Cas d^signant 1 agent dans certalnes 
langues comme le basque .... These brief definitions are of course 
oversimplified; in spite of the strong association of morphological 
ergatlve case marking with the agentive role in the so-called ergatlve 
languages, the two are far from fully co-variant. 


AXMANOVA, A. S. 1966. Slovar' lingvis tidesklx terminov. Moscow: Sov. 

enclklopedija. [Translation mine, CVC] 
BURZIO, L. 1981. Intransitive verbs and Italian auxiliaries. MIT Ph.D. 

thesis, ms . 
KEYSER, S. J. and T. Roeper. 1984, On the middle and ergatlve 

constructions in English. Linguistic Inquiry 15. 381-416. 
MAROUZEAU, J. 1951. Lexique de la termlnologie Unguis tique . Paris: 

Geuthner . 
PESETSKY, D. 1982. Paths and categories. MIT Ph.D. thesis, ms. 
ROZENTAL' D. E. and M. A. Telenkova. 1972. Spravo<fnik llngvis tidfeskix 

terminov. Moscow: ProsvegdTenie . [Translations mine, CVC.) 

studies in the Linguistic Sciences 
Volume 15, Number 2, Fall 1985 


Herbert S. Coats 
University of Washington 

This paper reviews briefly several analyses that have been 
proposed for palatalized consonants in Russian: those of the struc- 
turalists, for whom palatalization was phonemic; of Theodore Light- 
ner, who proposed an abstract analysis in which consonants become 
palatalized before front vowels j of other linguists who have re- 
tained Lightner's palatalization rule in less abstract phonological 
systems that typically posit a rich inventory of underlying front 
vovjels, such as /i e a b u/; and finally of Gilbert Rappaport, who 
retains the palatalization rule in an analysis in which the under- 
lying front vowels are /i e/. This approach is considered here the 
most promising because it allows for an analysis that is concrete 
and psychologically plausible, one constrained by the transparency 
condition. Several problems that arise in this framework are dis- 
cussed, including the exceptional behavior of /s z c/, the relation- 
ship between /i/ and /i/, and the behavior of velars. This analysis 
provides a novel but straightforward account of the shift of /velar 
♦ i/ to [palatal ♦ i] . 

In the traditional structuralist view of Russian phonology, a view that 
dates back to Rcnan Jakobson and many of his students, palatalization was 
considered phonemic for most consonants, so that forms like 1 jubit' 'to love' 
and pjat' 'five' were analyzed as follows: 

/I'ub'it'/ ■» Q'iib'it'] 
/p'at'/ -» rp'at'] 

The only rule that applies in these examples is one that fronts vowels be- 
tween palatalized consonants. In classical phoneirdc theory there was a 
simple and direct relationship between phonemic and phonetic representations, 
such that each can be deduced from the other, the so-called biuniqueness re- 
quirement. Alternations of palatalized (soft) and nonpalatalized (hard) 
consonants were described in the morchophonemic component. This was the 
generally accepted view of palatalization until Ted Lightner appeared on 
the scene. 

Lightner, impressed by Halle's famous argument against biuniqueness 
(Halle, 1959, pp. 19-23), rejected the phonemic level and worked out a de- 
tailed phonology of Russian that attempted to account for the whole range of 
data described previously in both the morphophonemics and phonemics. Light- 
ner's synchronic phonology mirrored the historical development of Russian: 
His underlying representations looked like Proto-Slavic forms, and his phono- 
logical rules looked like historical sound changes that took place in the 
evolution of Russian. Lightner's analysis of the two forms cited earlier, 
for example, would be roughly as follows (see FTP, pp. 57, 201-2): 

/leub-i-t5/ •¥ A'eub'-l-t'X/ -♦ /I'ub'-i-t'I/ .» A'ub'-i-t'/ -» tl'ub'-i-t'l 
/pint-i/ .f /p'int'-i/ * /p'ant'-I/ -» /p'at'-l/ -* /p'at'/ ■» tp'at'l 

The first rule above is one that palatalizes consonants before front vowels. 
This is a familiar and natural phonological process, one that is described in 
the SPE feature system as a kind of assimilation: Consonants are marked 
r+high, -back! when they occur before vowels marked [-back]. In Lightner's 
analysis later rules sanetimes alter these front vowels — they shift to back 
vowels or are dropped altogether, so that the vowels that caused the palatal- 
ization are not present in the phonetic representation, as in the examples 
above. Neither the opacity of this analysis nor its abstractness ever both- 
ered Ted very much. He was attracted by the rigor, explicitness, and, in a 
sense, the simplicity of this approach to phonology, and throughout his ca- 
reer he pushed it, with great consistency and perseverance, to the limit. 

It might be expected that a phonology that recapitulates historical 
sound change might run into complications in describing phenomena that arose 
by analogy rather than regular sound change, and indeed Lightner's analysis 
of palatalization provides instances of this. For example, in Old Russian 
there were i-stem nouns like kost' 'bone', which had the prepositional plural 
form kostixu, in which the t was palatalized due to the following front vowel. 
Then the a-stem ending axu spread by analogy to other nouns, yielding kostjaxu , 
with the original soft t retained before a new back vowel. In his analysis 
of this form Lightner posited underlying kost-axu , and now he had the problem 
of accounting for the palatalization of the stem-final t. In order to describe 
this and other soft-stem ncuns, Lightner marked their stems -HARD , and he 
expanded the palatalization rule to apply to the final consonant of a stem so 
marked (FTP, pp. 277-90). Note that this analysis accounts for the hard t 
in the diminutive kosto^ka, because the t in kost is not in stem-final posi- 
tion in this form. 

Lightner's basic line of cinalysis, that consonants are palatalized before 
front vowels, has considerable appeal not only because it is a natural phono- 
logical process, but also because it accounts for innumerable alternations of 
hard and soft consonants in Russian, as in stol 'table' and the following 
forms that contain this root: stola , stolu , stoli, stolovij , all with hard 1, 
as opposed to stole, stolik , which contain soft 1 before a front vowel. In 
addition, there is a clear restriction of the distribution of hard consonants 
in Russian: With a few exceptions that will be discussed below, hard conso- 
nants do not occur before front vowels. A palatalization rule could account 
for this fact. It is therefore not surprising that a number of linguists who 
are unwilling to accepxt Lightner's degree of abstraction are attracted by the 
idea of predicting palatalization before front vowels. Among such linguists 
are Lunt (1975), DeArmond (1975), and Bratkowsky (1980). Although each has 
his own approach, in general they argue that the inventory of phonemes in 
Russian contains the front vowels /i e a 6 ii/ and that these vowels cause 
palatalization of a preceding consonant. Their analysis of Ijubit' and pjat' 
is roughly the following: 

Aiibiti/ H [I'ub'it'l 
/patl/ ^ [p'af] 

It is clear that the vowels /vi i a/ will palatalize the preceding consonants. 
It is less clear how the t becomes palatalized in these forms, and each of 
the linguists cited above solves this problem in his own way. In the examples 
above I have followed Lunt's proposal, taken from Lightner, that the under- 
lying form contains a lax X which palatalizes the preceding consonant and then 

drops out. In order to account for forms like kostjax , Lunt marks their stems 
♦FRONT , which, like Lightner's -HARD , has the effect of producing palatalized 
t in this form, although in a way different from Lifhtner's. 

Although I sympathize with the aims of this approach and in particular 
with the desire to retain the palatalization rule in a framework less abstract 
than Lightner's, I find this approach still a bit too abstract for my taste. 
Like the structuralists, I believe tliat there should be a simple and direct 
relationship between phonemic and phonetic representations, a relationship 
defined not in terms of biuniqueness, but rather transparency. By trans- 
parency I refer to the requirement that the phonological entities — the 
segments and boundaries — which condition a phonological process be present 
on the sxirface, in phonetic representations. I believe that it is the prov- 
ince of phonology to describe only those phenomena — phonological alternations, 
restrictions on the distribution of sounds, etc. — which can be character- 
ized in terms of conditioning factors that are readily accessible, factors 
present at the phonetic level. The transparency constraint appeals to me 
because there is at least a possibility that phonemic representations sub- 
ject to this constraint may have psychological reality. It seems less likely 
to me that highly abstract underlying representations have psychological re- 
ality, but of course, as Lightner would point out, this is idle speculation, 
given our present knowledge of such iriatters . 

However, many of the phonological processes described by Lightner (in PTP, 
for example) happen to conform to the transparency condition. Among them are 
regressive voicing assimilation of obstruents, the devoicing of obstruents in 
word-final position, the reduction of unstressed vowels, the fronting of vow- 
els between palatalized consonants, etc. His description of the palataliza- 
tion of consonants, of course, does not conform to this constraint, nor do those 
proposed by Lunt, DeArmond, and Bratkowsky, because the phonological factors 
that condition the palatalization of t in p.jat' are not present at the phonet- 
ic level. I believe the same is true for kostjax , which I would expect to 
contain the ending /ax/, as in stolax , although Lunt posits the ending /ax/ 
here, a variant of /ax/ that occurs with stems marked ♦FRONT . This complica- 
tion could be eliminated and the transparency condition observed by positing 
underlying soft consonants. Our two forms now have the phonemic representa- 
tions /kost'-ax/ and /pat'/ or /p'at'/. But if Russian contains underlying 
soft consonants, then there is no reason to posit /pat'/ rather than /p'at'/, 
particularly at the expense of including in our inventory of phonemes the 
vowels /a ii 6/, phonemes which are no longer necessary and which, I suspect, 
have little psychological validity. 

Rappaport (I'^Bl) has suggested that Russian contains the vowel phonemes 
/i e o u i a/ and that the two front vowels here, /i/ and /e/, cause palatal- 
ization of preceding consonants. Palatalization in other positions is not 
phonemic. This analysis is consistent with the transparency condition; it 
retains the rule that palatalizes consonants before front vowels, which is a 
natural phonological process; it accounts for many alternations of hard and 
soft consonants, as in the examples containing the root /stol/ cited earlier; 
and it accounts for the restriction on the distribution of hard consonants 
that was mentioned earlier. In this analysis the palatalization of conso- 
nants is comparable to voicing in obstruents. Most obstruent phonemes come 
in pairs such as /p b/, /t d/, etc., but in certain positions, such as at the 

end of the word, this contrast is lost because the rule of final devoicing 
shifts voiced obstruents to voiceless ones in this position. Similarly, 
most consonants come in pairs such as /t t'/, /s s'/> etc., but before a 
front vowel this contrast is lost because the palatalization rule applies in 
this position. In the remainder of this paper I shall explore several prob- 
lems that arise if one takes this analysis seriously. 

The first problem concerns the few consonants which are not paired with 
respect to palatalization. These are /c ' j/» which are always soft, and 
/z i c/, which are always hard. We account for these facts by saying that 
these phonemes are appropriately marked for the features thigh] and [back] , 
possibly by redundancy rules. The problem is that /z s c/ do not become pal- 
atalized when they occur before front vowels, as in no^e [naze] 'knife (prep- 
ositional singular)', phonemically /noz-e/. Now /z i~c7 differ from other 
hard consonants in that they are strongly velarized, while the others are 
only slightly velarized — see Avanesov (1972, pp. 3U-hU) . That is, /z s c/ 
are marked f+high, ♦back], while the others are [-high, ♦back!, in contrast 
to plain consonants, which are [-high, -back] — see SPE (pp. 305-8), where a 
different interpretation is given for consonants marked [-high, ♦back] . I 
propose that the palatalization rule be restricted to apply only to conso- 
nants that are specified [-highT. This now excludes /z § c/ from xindergoing 
the rule, and it does so in a way that seems fairly plausible. It is not un- 
reasonable that strongly velarized consonants might resist palatalization. 

This treatment of /z s c/, however, leads to a problem with the velars 
/k g x/, which do undergo the palatalization rule, as in ruke [ruk'e] 'hand 
(prepositional singular)' and deduskin Id ' edusk ' inJ 'grandfather's', which are 
derived from /ruk-e/ and /dedu§k-in/. The velars are of course specified 
[♦high, ♦back], and so the palatalization riile as presently formulated will 
not apply to them. Note that the shift of /k g x/ to /k' g' x'/ does not 
really involve palatalization, but is rather the change of velars to palatals. 
That is, the primary point of articulation changes when this shift occurs, 
while palatalization per se involves a change in secondary articulation only. 
These considerations lead me to suggest that the shift of /k g x/ to /k' g' 
x'/ should be described by means of a rule quite independent of the palatal- 
ization rule, a rule that applies only to velars. The proposed rule is given 


[-back] / 1*"°^ 1 
I -back J 

The distribution of [k' g' x'] gives some additional motivation for analyz- 
ing them differently from the palatalized consonants. With very few excep- 
tions, most of them foreign borrowings, these sounds occur only before the 
phonemes /i e/, and of course in these positions they are derived from under- 
lying /k g x/. Consequently the phonemes /k' g' x'/ are restricted to the 
few exceptional forms referred to above. As we have seen in examples like 
1 jubit' and pjat' , the palatalized consonants are not restricted in this way. 

I proposed above that the palatalization rule be formulated so as not 
to apply to /z s c/. The rule must also be formulated so as not to apply 
across the pre fix -preposition boundary )f . The phonemic representations 

/s#ekranom/ ' the screen', /s#ivanom/ 'with Ivan', and /s^iprat'/ 'to 
play', for example, yield [s#ekranom] , [si^ivanoml and Is^lgrat'] , with no 
palatalization of the initial s in these forms. (The backing of /i/ to Til 
in these examples will be discussed below.) The proposed palatalization 
rule for Russian is given below. I adopt here the convention, which as I 
understand it is proposed in SPE (p. 6?), that rules automatically apply 
across a formative boundary, but they apply across larger boundaries such 
as # only if these boundaries are specified in the rule. 

(■•fcons] r+high] . r+voc ] 

L-highJ L-backJ ' l-backJ 

This rule will not apply across #, and neither will the rule for velars given 
on the preceding page, which is the correct analysis in view of forms like 
/toi'ekranu/' 'to the screen' and /k#ivanu/ 'to Ivan', which yield phonetic 
noi'ekranu] and Ck#ivanul . 

T would now like to discuss briefly the relationship between /i/ and the 
high, back, unrounded vowel /i/ . The phonemic status of the latter has been 
a controversial question in Russian phonology for a long time. Most recent 
studies, such as those of Bratkowsky (1980) and Rappaport (1981 ), provide 
new evidence that /i/ is indeed an independent phoneme, and in this paper I 
have assumed, and shall continue to assume, that it is. There are, however, 
instances in which /i/ is realized phonetically as Cil and vice versa. We 
have seen, for example, that the underl>'ing nominative plural ending in 
nouns is /i/, as in /stol-i/, which yields Cstali] . After a palatalized 
consonant, however, this /i/ is realized as [il — an example is tlos'i] 'elks', 
derived from underlying /los'-i/. Thus we need a rule that shifts /i/ to Ci] 
after a palatalized consonant. If we compare two dimunitive forms like 
[stol'ixi and Inozik] 'knife', derived from /stol-ik/ and /noz-ik/, we see 
that we also need a rule that shifts /i/ to Til after a hard consonant. These 
two rules can be collapsed by using the alpha convention: 

[+VOC 1 
.high -. r-backi / l:ill^] (#) 

The optional # in this rule allows it to apply in forms like /s#ivanom/ which 
were discussed above. 

There remains one problem to be discussed. It turns out that /i/ shifts 
to [i] after a velar, as in nominative plural Ldux'il 'spirits', derived from 
/dux-i/. Within the framework developed above there are several ways to handle 
such forms. We could posit a rule that shifts velars to palatals before /i/, 
and then the rule given above would shift the /i/ to [i] ; that is, the deri- 
vation would proceed as follows: /dux-i/ •♦ /dux'-i/ -♦ Idux'i] . Or one could 
introduce a rule that fronts /i/ after a velar, and then the rule given earlier 
that shifts velars to palatals would apply — i.e., /dux-i/ •♦ /dux-i/ -» Fdux'iK 
Finally, we could say that two rules already formulated above, the one that 
shifts velars to palatals and the one that fronts /i/ to Til , apply simulta- 
neously, each producing the segment that functions in the environment cf the 
other. Notice that the phonetic representation [dux'il is transparent with 
respect to both of these rules. The rule that shifts /x/ to Ex'] requires a 

following front vowel, which is present in Cdux'i], and the rule that fronts 
/i/ to li] requires a preceding consonant marked [-back], which is also 
present in [dux '11 I prefer this last analysis because it is the siir.plest — 
it does not require a new rule — and it is consistent with the transparency 
condition. It also avoids unnatural intermediate representations like 
/dux'-i/ and /dux-i/. 

A final comment about rule ordering. With a single exception, none 
of the rules formulated in this paper are extrinsically ordered. They can 
apply in random order or, as suggested above, they can apply simultaneously. 
The exception involves the palatalization rule, which must apply before the 
rule that shifts /i/ to tilj underlying /stol-ik/ must yield [stol'ik], not 
[stoltk] . I consider this an inadequacy of the present analysis. A possible 
solution might be to posit underl>'ing plain consonants rather than what I 
characterized above as slightly velarized ones. That is, such consonants 
will be marked I-high, -back] rather than I-high, ♦back]. The derivation of 
[stol'ikl now works out all right, but we still have to account for forms 
like [s#igrat'l and for the fact that the hard consonants in Russian are 
slightly velarized. I end this paper with a challenge to the reader to 
find a solution to this problem. 


AVANESOV, R. I. 1972. Russkoe literatumoe proisnosenie. Moscow: Pro- 

BRATKOWSKY, Joan G. I98O. "The predictability of palatalization in Russian". 
Russian Linguistics hO 329-336. 

CHO>EKT, Noam and Morris Halle. 1968. The sound pattern of English. New 
York : Harper and Row. 

DeARMOND, Richard C. 1973. "The Russian phoneme revisited". The proceed- 
ings of the 2lith annual pacific northwest conference on foreign lan- 
guages. Corvalis: Oregon State University Press. 

HALLE, Morris. 1959. The sound pattern of Russian. The Hague: Mouton & Co. 

LIGHTNER, Theodore M. 1972. Problems in the theory of phonology. Edmonton: 
Linguistic Research, Inc. 

LUNT, Horace G. 1975. "Phonological and morphological units in teaching 
Russian". Slavic and East European Journal 19:1 7U-89. 

FTP = Lightner, Theodore M. 1972. 

RAPPAPORT, Gilbert. I98I "Distinctive and redundant contrasts in Jakob- 
sonian phonology: a review article". Slavic and East Eioropean Journal 
25:3 9I4-IO8. 

SPE = Chomsky, No<im and Morris Halle. 1968. 

studies in the Linguistic Sciences 
Volume 15, Number 2, Fall 1985 

Alice Davison 

Non-finite clauses of various kinds may lack overt subj- 
jects, but the referent of the null subject is supplied 
by an antecedent, or controller. Control may by obliga- 
tory or optional, allowing an arbitrary interpretation. 
For Hindi-Urdu, the control properties of a non-finite clause 
are determined by its syntactic relation to the matrix clause. 
VP constituents are clauses which are obligatorily controlled, 
others are optionally controlled if there is an available 
antecedent. Obligatorily controlled null subjects lack case; 
they cannot be NPs originating in positions which receive overt 
postpositions, such as the dative-accusative which marks exper- 
iencers or patients. The different control properties of non- 
finite clauses and the restrictions on null subjects suggest 
that Hindi-Urdu can be regarded as a configurational language, 
in which phrase structure relations are relevant to rules of 
grammar, and that non-finite clauses should be represented as 
having a syntactic subject position. 

Languages like Hindi-Urdu appear to have relatively free word order in 
surface syntax, and to lack syntactic rules which reorder NPs, changing 
grammatical relations. For these and other reasons Hindi-Urdu might be 
considered a nonconfigurational language, in which grammatical relations 
are not encoded in structural relations characterizing phrase structure 
trees. Instead, on one view, case marking and grammatical relations might 
be assigned by verbs to constituents designated in non structural ways as 
their arguments, by lexical frames (cf. Farmer (1985) for example). Yet a 
closer examination of the specific patterns of case marking and the rules 
for the control, or assignment of reference, to null subjects in Hindi-Urdu 
reveals features of the syntax which lend themselves to description in 
configurational syntactic terms. It will be proposed in this paper that 
case-marking and control in a language like Hindi-Urdu, with some freedom 
of word order and extensive case-marking, can be considered a 
configurational language, and that syntactic encoding of grammatical 
relations is a general principle from which the restrictions on case 
marking and control follows natural consequences. 

The discussion will be centered on NPs with dative-accusative 
case-marking. These NPs have a dual nature, with both object and subject 
properties. For example, passive object NPs may retain dative-accusative 
case in S-Structure; experiencer NPs always have dative marking. Yet both 
kinds of dative-marked NPs also have properties of subjects, as in other 
languages such as the Quechua languages (Hermon (1981), (1985)) and Nepali 
(Wallace (1985)). Dative NPs may undergo Raising, be marked as subjects in 
Exceptional Case Marking contexts, and be the antecedent controlling 
reflexives and PRO, which require subject antecedents. This paper takes 
the position argued for in Davison (1969, 1985), and the analyses by Hermon 
and Wallace that NPs with surface non-subject marking are subjects at 
another level of derivation. Within the Government and Binding framework 
of Chomsky (1981), NPs get case marking in S-Structure by virtue of their 


configurationally defined roles as arguments of the verb in their clause, 
but acquire subject properties by being moved in the derivation of LF to 
subject positions, where they meet the requirements for binding of 

This analysis of dative-marked NPs makes possible some interesting 
statements about the role of case-marking in this language, and about the 
properties of case-marked subjects, including dative subjects. When 
control of PRO is obligatory, case-marked subjects are prohibited. These 
facts are of some interest in showing the consequences of language 
specific features for the general description of phenomena like the 
coindexing of PRO and its antecedent. They also point to specific 
grammatical properties of null subjects, a matter which is controversial 
and has led to the treatment of infinitives as bare VPs (Gazdar et al 
(1985)) rather than as clauses with empty subject positions. Finally the 
analysis proposed here will suggest that the level of syntactic description 
Logical Form has some syntactic functions which parallel the functions of 
the derivation of S-Structure in other languages. 

An example of control in the conjunctive participle construction is 
given in (1) below. The conjunctive participle is marked by the affix -kar 
on a tenseless verb. 

1) bahut baar woo. 1 PRO. saRak bhuul - kar J bhaTak gayaa 
many times 3psg road forget conj.part be-lost go-perf 

' Many times PRO forgetting the way he got lost' (T.Q. 42) 

2) mistrii. - nee [_ PRO. /* baniyaa. kal aa - kar l 
carpenter erg shopkeeper yesterday come CP •* 

sanduuq banayaa 
box build-perf. 

'The carpenter having come yesterday made the box'; *'The 
carpenter made the box the shopkeeper having come yesterday' . 

The sentences in (1) and (2) containing the conjunctive participle in 
a subordinate clause illustrate the control and case-marking properties 
which will be the main topic of this paper. It will be assumed that 
sentences like (1) and (2) contain a main clause, with all its arguments 
potentially realized as lexical items. These two sentences also contain a 
subordinate clause which has the otherwise tenseless verb marked with - kar , 
meaning roughly 'perfective' The subject of the conjunctive participle 
clause is always null, and always coindexed with the subject of the main 
clause. This contruction is a case of obligatory syntactic control, 
contrasting with constructions of other syntactic types which allow the 
subject to be lexically realized, or optionally controlled by an 
antecedent. In such cases there may be no syntactic antecedent, allowing 
for discourse control of a null subject which has essentially an arbitrary 
interpretation . 

Case in Hindi-Urdu is assigned to NPs by the verb or verbal complex 
which governs them. Direct objects are optionally marked with the 
dative-accusative post-position -koo, though some verbs require specific 


postpositions such as - par 'on^ or - see 'with', which are obligatory. 
Subjects have either no case marking at all, and appear in the ^direct^ 
form if they are inflected words, or else get the ergative marker - nee if 
the verb is transitive, lexically marked for ergativity and if the aspect 
of the clause is perfective. Infinitive clauses are never inflected for 
perfect aspect per se , so infinitives never assign ergative case to their 
subjects, whether overt or null. Dative subjects are a somewhat special 
case, but are obligatorily marked by their c-commanding verb with - koo in 
a way to be described in further detail in the paper. In examples (1) and 
(2), the case-marking appears only on the main clause subject, and is 
determined by the main clause verb. In the sentence in (1), the main verb 
is intransitive, so the subject has null case, while in the main clause in 
example (2) , the subject has ergative marking because the main clause verb 
is perfective and transitive; the verb of the conjunctive participle 
clause is intransitive, and would mark its subject with the null or direct 
case if it were overt. The verb of the conjunctive participle in example 
(1) is grammatically transitive, taking an object, but does not assign 
ergative marking to its subject. 

The grammatical rules of Hindi-Urdu allow null subjects in subordinate 
clauses, which may have syntactic antecedents, or must have syntactic 
antecedents; their control properties are determined by syntactic 
relations, as we shall see in more detail later in the paper. Besides the 
conjunctive participle construction, other syntactic constructions which 
require PRO subjects to be controlled either by subject or object syntactic 
antecedents include infinitive object clauses which are complements of 
verbs like caahnaa ^want', kahnaa ^say' and siikhnaa ^learn'. Examples (3) 
and (4) illustrate PRO as a complement subject controlled by a subject or 
an object. 

3) pro.] PRO. kheelnaa J jaantee nahTi — iPRO kheelnaa j 
play-inf. know-2ppl not play-inf. 

siikhoo .fam. Porizka (1963:340) 'You don't know how to 

play (chess); learn to play' 
4) us.- nee surendra .- koo L PRO. 

saath calnee -kee 

3psg erg S "^ DAT. "^ with go-inf. of 

liyeej kahaa 

sake say-perf. Ibid p. 431. 'She asked Surendra to walk 
with her' . 

In the sentence in (3) , PRO is controlled by the syntactic subject of the 
sentence, which is not lexical, but its properties of being second person 
familiar, or tum are recoverable from the person and number inflections on 
the verb. The antecedent of PRO in the sentence in (4) is surendra , whose 
postpositional marking shows that it is an indirect object of the main verb 
kahaa . 

Syntactic control by a subject or indirect object antecedent is 
required for PRO in these cases. The syntactic contexts where control of 


PRO subjects is obligatory include the infinitive objects of verbs and the 
tenseless clauses marked by the conjunctive participle affix -kar, both 
constituents of VP. As the complement clauses are tenseless, the subject 
could not be assigned ergative marking even if it normally has such a 
marking in a main clause. Hence there is no reason to assume that the PRO 
subject in these examples would be assigned any case. It is 
indistinguishable from the lexical subjects which have the ^direct' 
morphological form rather than the oblique form of inflected nouns etc., 
and no postposition. The pattern of obligatory syntactic control with no 
case-marking illustrated above will now be contrasted with another pattern 
in the following examples. Here, control is pragmatic, PRO is not 
obligatory, and case is assigned to the NP in subject position. 

The syntactic contructions with the latter set of properties includes 
subject infinitives, participial adverbial clauses and modifier clauses. 
These are illustrated in (5)-(7) below: 


1 PRO , pyaar karnaa J zindagii /PRO jiinaa hai 
'• love do-inf. life live-inf. is 

^PRO ^ to love is life/ PRO to live', 
arb X 

6)1 PRO. deer soonee - see J mujhee. sirdard hoo jaataa 
late sleep-inf .from Ipsg-DAT headache be go-impf 

hai. ^I get a headache from going to bed late'. 

7) kaun sii baat [ PRO . karnee -kaa J hai, aur kaun sii 
what like matter do-inf .-obi .of is and who like 

nahii? "What should one do and what should one not do?' 

(Porizka (1963:338)) 

In the sentences in (5) and (7), the PRO subject in the first, or subject 
infinitive has no syntactic antecedent, and so refers arbitrarily. 
Information in the discourse context may be used to fix the reference. The 
PRO subject in the second infinitive clause in (5) is controlled by the PRO 
of the first clause; that is, it does not have independent arbitrary 
reference. The PRO subject inside the adverbial clause in (6) is 
coreferential with the dative subject of the main clause; as we will see 
shortly, clauses with infinitival or participial morphology and adverbial 
function allow optional control. 

The clauses which have PRO subjects of the types just illustrated all 
allow overt lexical subjects which do not have to co-refer with any other 
part of the sentence. Counterparts to sentences (5) - (7) are given in (8) 
-(10) below: 

jsarfraaz kaa kyaa karnaa"! ajiit 
IS of what do-inf . -I strane 

I) I sarfraaz kaa kyaa karnaa] ajiib ki i baat hai? 

ange of matter is 

"What is it strange for Sarfraz to do? 


9) 1 meeree kyaa karnee - see | pitaa -koo gussaa aayaa? 

1 Ipsg-of what do-inf .-obi .from father-DAT anger come-perf 

^Father got angry on account of my doing what?' 

10) kaun sii baat fmeeree karnee kii Jhai, aur kaun sii 
who like matter Ipsg-of do-inf .obl.of is and who - like 

tumharee? 'What should I do, and what should you? 

In each of the three sentences above, the PRO subject which occurred 
in (5) -(7) has been replaced by a lexical NP with genitive case, which is 
normal for NPs referring to animate beings. Subject NPs referring to 
inanimate objects have case: 

1 1 ) 1 yahaa - see landan -tak bhaaii - kee/ciTThiiJB pahucnee- 

here from L. to brother of letter reach-inf. 

in J 

kitnee din lagtee hai? 
how-many days strike-pl is-pl 

'How many days does it take for brother/ for a letter to reach 
London from here?' (Porizka (1963: S'JS)) 

As the example in (11) shows, the subject of the adverbial clause may have 
either genitive or null case, depending on the properties of its referent. 
In all these syntactic contexts, PRO and control of PRO are not obligatory, 
and there is a case, the genitive, associated with the complement subject 
position. There may be null subjects in adverbial clauses which are not 
syntactically controlled. This is the case in the example in (12): 

12) (_ PRO caltee caltee J andheeraa hoo gayaa 

walk-impf-walk-impf darkness become go-perf 

'As (he) went along, darkness fell'. (Porizka (1963:359)) 

Here the subject of the modifying clause is null, and certainly not 
identical in reference to the subject of the main clause andheeraa 
'darkness'. Of course, there are other sentences in which the PRO subject 
of the adverbial could be controlled by the subject of the main clause. 
The reference of the subject of the participial modifying clause can be 
determined by the discourse context, and is not restricted by syntactic 

It was mentioned earlier that there are some reasons to believe that 
dative NPs, both direct and indirect objects, have subject properties of 
various kinds. Two kinds of evidence will be given below in the examples 
in (13) and (14): 

13) too kumar.-koo kaisaa lageegaa, [f^O- unhee . 
then K. ^ DAT. how strike-fut. ■^3pp1.D/Jt. 


deekh - kar? ■" J 
see Conj. Part, 

"Then how will it strike Kumar, PRO having seen them?' 
(Mohan Rakesh, Antaraal ) 

The subject position of the conjunctive participle clause is obligatorily 
null. This null subject of the verb deekh-kar must be controlled by a 
subject in the matrix clause, in this case kumar-koo , which is marked with 
the dative postposition as the experiencer complement of the verb lageegaa . 
This verb is subcategorized for an experiencer. It is proposed in Wallace 
(1985) and Davison (1985) that experiencer NPs are verb complements in 
syntactic structure, but are moved in the derivation of Logical Form to 
subject position. The requirements for control of PRO would be relevant to 
LF, not to S-Structure (as proposed for Quechua by Hermon (1985)). 

A similar account is proposed in Wallace and Davison for passive 
objects. Object marking by - koo is not obligatory, so that some objects 
have the same case, or rather absence of case, whether they are in active 
or passive clauses. If a direct object has the dative postposition, it 
does not lose it as a result of being in a passive sentence and in 
being moved to subject position with absorption of object case. Hence the 
passive construction may be analyzed as having no instances of move alpha 
in the derivation of S-structure. Case is assigned, or not, to the direct 
object on the basis of definiteness and animacy. In the derivation of LF, 
the object NP may be moved to subject position, retaining its case marking. 
The sentences in (T^) provide some evidence for the subject properties of 
passive objects: 

1'4) a. raam - koo pulis - see piiTaa gayaa 

R DAT police- by beat-perf .go-perf 

"Ram was beaten by the police.' 

koo pulis -see piiTee jaatee huee J 
DAT police by beat-perfgo-impf be-obl. 

mai- nee 



I erg. 



see-perf . 

"I saw Ram being beaten by the police.' 

c. mujh-see \ raam - koo pulis- see piiTee jaatee huee "| 
I-obl.-by R DAT police- by beat-pf .go-impf be-obl. 

nahii deekhaa gayaa. 
not see-perf go-perf. 

"I couldn't bear to see Ram being beaten by the police'. 

raam-koo in (14) a. is the object in a passive sentence, marked by the 
perfective participle on the main verb and the auxiliary jaa "go*. The 
dative marking is retained in (14) b, in which the passive clause is the 
object of the verb deekhaa , a verb which requires that its complement has 
an imperfective oblique verb with an auxiliary huee "become*. The subject 


of the complement clause must be lexical, and must have dative marking, a 
counterpart of Exceptional Case Marking. As in similar cases in other 
languages like English, the complement subject with object marking has the 
properties of an object with respect to the verb of the matrix clause. 
Hence we find that in (14) c, the dative NP is also the object of deekh 
(unless we want to treat the sentence in (14) a. as an instance of an 
impersonal passive). But in any case, the very same NP marked as an object 
satisfies the requirement for a complement subject marked with the dative 
postposition, a requirement which is the property of object clauses of the 
verb deekhnaa ^see'. 

NPs marked as datives have a dual nature. They receive case as 
objects, direct or indirect; they are syntactically complements of the verb 
which assigns their case, and grammatical role. These features are ones 
which are assigned to lexical items, and are instantiated in specific 
sentence structures, as properties of complements of tokens of a lexical 
item. But nothing in the S-Structure representation of sentences with 
experiencer predicates or passive verb phrases marks NP - koo as a subject. 
Its case marking assigned by the verb reflects its grammatical role, or 
theta role, of goal or theme. But these NPs also have subject properties in 
being well-formed antecedents for reflexives and controllers of PRO. Their 
subject properties are not reflected in the case marking which they retain 
even when serving as subject antecedents. If dative NPs are also subjects, 
then they are subjects by virtue of occupying a configurationally defined 
subject position without having their dative case marking be absorbed by 
passive morphology (cf Chomsky (1981)) or be replaced by a case marker 
assigned by INFL, or - nee . Subjecthood is not a grammatical role, and in 
any case, the dative NP already has a grammatical role, as patient or 
experiencer. Case does not get absorbed or assigned by inflection, since 
the dative case is just the case assigned by the verb to its complement. 
Hence the subject properties of dative NPs must be represented in some 
other way. 

One method of representing subject properties is to say that the dative 
NP is moved from its origin inside VP to a null subject position. There is 
no principle which makes it obligatory for the NP to move, no case filter 
for example. If it does not move, the resulting construction is simply 
impersonal, with a null subject rather than an expletive rt, etc. Sentences 
such as pitaa -koo kroodh aayaa ^father got angry' could equally well be 
interpreted as examples of an impersonal construction as of a dative 
subject construction. The tree structure showing movement is as follows: 





If pitaa-koo is moved to the null subject position, it takes on the subject 
properties relevant for control and the binding principles, which must hold 
at LF in this language. It can be the antecedent for a reflexive or a null 
subject PRO in an infinitive, participial or conjunctive participle clause 
(cf. Pandharipande (1981) for discussion of the subject properties of the 
analogous passive cases; see Davison (1969, 1985) for more examples of 
dative subjects.) 

A configurational representation has been proposed in (15) for the 
subject properties of dative NPs. It is also possible to define in 
configurational terms the difference between non-finite clauses which 
require null subjects and syntactic antecedents, and those with optional 
null subjects and arbitrary readings for them in the absence of a syntactic 
antecedent. Examples of obligatory and optionally controlled null subjects 
are illustrated in (16): 


/ v-ini\ 
i- V-kar5 












iO / 

/ 1 
/^ V 












The tree in (16) a. represents object infinitive complements as well as 
conjunctive participial adverbials as constituents of VP. These complement 
clauses have obligatorily controlled PRO subjects. The controller must be 
a subject which c-commands and precedes the PRO subject. The geometrical 
relation between the complement clause in the VP and the subject position 
is such that there is necessarily a subject position which both c-commands 


and precedes the clause containing the null NP. The opposite is the case 
in the tree in (16) b. The complement clause, a subject complement or 
adverbial clause, precedes and c-commands any possible antecedents. If the 
complement clause is a subject, then there is no subject for it and its 
subject to be coindexed with without violating the i-within-i restriction 
(Chomsky (1981)): 

17) X 

. . . Y. . . . 


Identical referential indices on both a subconstituent and the entire 
constituent are prohibited. (Cf. Manzini (1983) for a proposal which 
relates the control of PRO to the binding of anaphors. There are some 
parallels in Hindi-Urdu between control of PRO and the binding of 
antecedents and reflexives, which cannot be discussed here.) In the case 
of adverbial clauses dominated by S and preceding the subject, an 
antecedent is available which does not violate i-within-i. But it is 
preceded and commanded by PRO. We will assume that no syntactic binding 
relationship is possible here, for the reason that the geometrical 
relationship between PRO and antecedent is wrong. It is interesting to note 
that overt pronouns cannot be followed by their antecedents and be 
syntactically coindexed (Kachru (1978)). Here PRO cannot be syntactically 
bound by its antecedent. But in such a case, PRO has the ^arbitrary' 
reading, allowing for discourse control. The preceding clause counts as 
prior discourse, allowing for optional control and none of the strict 
requirements that the controller be a subject, etc. 

In (18), the two kinds of relations are contrasted. The version in 
which the subject clause has a null subject is grammatical, with the 
* arbitrary' reading. The version with the null NP in a conjunctive 
participle is ungrammatical , since there is no antecedent, yet clauses 
which are VP constituents are obligatorily controlled: 

18) a. I PRO aisaa karnaa 1 deesdroohii hai 
I such do-inf . J treason is 

rnaa ] 
-inf. J 

^PRO to do such a thing is treason'. 

"I ^ 

, aisaa kar-kee deesdroohii hai 

such do-CP. J treason i£ 


^??PRO having done such a thing is treason'. 

The contrast between these two nearly identical sentences is shown in (19) 


X \ 














1 1 



aisaa karnaa 

such do-inf, 



deesdroohii hai 
treason is 

aisaa kar-kee 
such do-Con j-Part. 

The two sentences (18) a. and b. are lexically very similar but 
syntactically different. The non-finite clause is a subject clause in one 
((18) a.) and a VP modifier in the other ((18) b.). The null subject in the 
subject clause receives the arbitrary reading because the only subject in 
its matrix clause which could be coindexed with it is the sentence itself, 
and this coindexing of NP* and S* would violate the i-within-i restriction 
(Manzini (1983)). The VP modifier clause must have a syntactic antecedent, 
which would normally be found in the matrix subject position. But this 
position is itself occupied by a null element, which is coindexed with the 
predicate nominal (or has no reference, like the expletive it.) 

whose const 
position is 
binding, th 
passive sen 

just proposed th 
and optional cont 
onally, we should 
The critical case 
ituents are like 
also a position 
of LF. In the cas 
e result is usual 
tences whose mean 

at subject properties of dative NPs should be 
y, and that the difference between cases of 
rol should also be represented 

now see how these two sets of relationships 
s involve complement clauses with null subjects 
those in (15). That is, the null subject 
to which a dative NP might move in the 
e of optional control, that is, non-syntactic 
ly well- formed (some exceptions being found for 
ing could be expressed with intransitive 



verbs). In the case of obligatory control, however, the null subject 

cannot be a subject which has case, a passive patient or dative 

experiencer . 

In the sentences in (19) below, a passive object is the subject of a 

conjunctive participle (19) a, or of a variety of participial and 

infinitival adverbials (19) b.: 

19) 2' 1 * ^^^- piiTaa jaa -kar ") laRkaa. calaa gayaa 

'- beat-perf. go- CP J boy go go-perf. 

^Having been beaten, the boy went off.' 

b. \ PRO. piiTaa jaanee-par /jaanee kee baad/ jaatee hii ~| 
L beat-perf go-on go-inf. after go-impf. only 1 

laRkaa. calaa gayaa 

boy go go-perf. 

^Having been beaten, the boy went off. 

c. ) PRO. piT - kar J laRkaa. calaa gayaa 

be-beaten CP boy go go-perf. 

^Having been beaten, the boy went off.' 

The conjunctive participle version is ungrammatical , while the others are 
not. There is nothing about the conjunctive participle itself which 
forbids coindexing of the antecedent with a semantic patient. The 
grammatical sentence in (19) c. demonstrates this, since it involves 
control of a patient NP which is the subject of the verb piT- ^be beaten', 
which is the intransitive counterpart of the verb piit- ^beat'. 

Similarly, object complements cannot contain null subjects which would 
be dative if they were overt. Only the intransitive counterpart of a 
transitive verb is possible, since it imposes only ^nominative' or null 
case on the patient — subject: 

20) a. mai . 1 PRO. luTnaa J nahii caahtaa huu 

1 1 
I be-robbed not want-impf. am 

^I don't want to be robbed,' 

b. * mai. PRO. luuTaa jaanaa J nahii caahtaa huu 
I ^ rob-perf. go-inf. not want am 

'l don't want to be robbed.' 

Object complements and conjunctive participle clauses may not have dative 
experiencer subjects: 

21) a. • mai. \ PRO. sird-dard hoonaa J nahii caahtaa huu 

I headache be-inf. not want am 


I don't want to get a headache' 

b. mai nahii caahtaa huu [ki (mujhee) sirdard hoo J 

I not want-impf,am that I-dat. headache be 

'I don't want to get a headache.' 

22) a.[»PRO. bukhaar aa-kar J baccaa. roonee lagaa 

fever come-CP child ory-inf. begin 

'Having gotten a fever, the child began to cry.' 

b. r baccee-koo bukhaar aanee-parj maa-koo taqliif huii 
child-obl.-dat. fever come-on mother-dat. worry was 

*The mother was worried when the child got a fever.' 

The null NP in (22) a., which gets dative case, is ungrammatical in the 
conjunctive participle construction, which requires control of null 
subject NPs. The infinitive counterpart in (22) b. does not require 
control, and does allow an overt dative NP, baccee-koo . Such a NP could be 
null, an instance of pro , if there is a discourse antecedent. The 
generalization which emerges from the above contrasts of well-formed and 
ill-formed sentences is the following: 

23) Controlled PRO does not have case. 

Whether or not PRO is syntactically controlled depends on the 
conf igurational relation between the clause containing PRO and a possible 
antecedent. If there is no antecedent, PRO will not be controlled, and 
will have the arbitrary interpretation. If there is an antecedent, but it 
follows the clause containing a null subject, control is not obligatory. 
If an antecedent precedes and c-commands the clause containing PRO, then 
PRO will be controlled. Even so, if PRO represents a subject position 
which might have case, it cannot be controlled. The null subjects in (19) 
b. and (20) b. are obligatorily controlled by a subject antecedent. But 
the null subjects also should receive case. They violate (23) , and the 
sentences are ill-formed. 

A contrasting case of a case marked null NP is shown in the sentence 
in (24): 

24) a.j_PRO , badlaa milnee kaa] mauqaa hai 

reward receive-inf of opportunity is 

^It's an opportunity PRO to receive a reward.' 

b. usee aachaa badlaa milaa 

3psg/DAT. good reward receive-perf 

'He got what he richly deserved.' 


The modifier clause in (24) a. contains a null subject which, as (24) b. 
shows, normally gets dative marking, and has dative subject properties 
(Davison (1969)). The modifier clause is in a configurational relation 
which prohibits coindexing with an antecedent — and in fact none is 
available. Hence the PRO subject has arbitrary reference and experiencer 
role; because NP is not controlled, it does not violate (23) even though it 
has dative case. 

Both control and dative subject properties can be expressed in 
configurational terms. Their intersection produces some unexpectedly 
ungrammatical results, which can be subsumed under the language-specific 
principle in (23). This principle defines the properties which some null 
subjects have, but not others. That is, some infinitival and participial 
subjects are controlled and others are not. Whether they are controlled is 
not a property of the clause in and of itself. Rather it is a property of 
the larger syntactic context in which the infinitive or participial clause 
is found. In a system of grammatical description which represents the 
subjects of infinitives and participles only in semantic translation 
(Gazdar et al (1985)) or in functional structure (Bresnan (1982)), if they 
are not overt, it would not be possible to generalize in this way about 
which coindexings with antecedents will be possible, and what grammatical 
role and case the subject may have. It has been proposed that null 
subjects have quite specifiable properties, even though they are not overt. 
Further, these properties depend on syntactic ocnf igurations, and cannot be 
completely account for as properties of lexical entries for verbs. 


An earlier version of this paper was presented at the May 1985 SALA 
meeting at the University of Michigan, Ann Arbor. I am grateful to Manoj 
Jha and Baber Khan for syntactic and semantic judgements of the examples 
included in this paper. 


Further examples of the conjunctive participle construction may be 
found in Davison (1980), Kachru (1981) and Abbi (1981). In this paper, I 
propose a different syntactic analysis from the ones in these papers. 
Although word order is variable, there is a tendancy for conjunctive 
participle clauses to follow nominative or ergative subjects, and to 
precede dative subjects in the matrix clause. This tendancy suggests that 
conjunctive participle clauses are constituents of VP, and not, as I 
proposed earlier, adjuncts of S. 


Not all complements of verbs etc. are lexically realized in every 

instance of use. Hindi-Urdu has null pronominals, or pro , whose reference 

is recoverable from discourse context. Lutz (1985) discusses a number of 

examples of null pronouns, whose reference is determined by discourse topic 

rather than by rules of grammar or morphological information. 


There is a small number of permitted exceptions to this 
requirement, which have not been fully explained either for Hindi-Urdu or 
for other South Asian languages with similar constructions, such as Tamil. 
In productive uses of the construction, there must be a null subject 
controlled by a matrix clause subject. 

In general, transitive verbs have -nee subjects. See Porizka 
(1963:425ff) for an overview of exceptions, such as bhuulnaa "forget', 
which is transitive but may have a nominative subject, and ciiknaa 
"sneeze', which is intransitive but requires a -nee subject. 

Note that this example illustrates the freedom of word order allowed 
in Hindi-Urdu. The conjunctive participle clause is right-dislocated. I am 
assuming that this word order reflects the possibilities allowed by 
Phonetic Form, without affecting the syntactic relations which are 
relevant for S-Structure and LF. 

In brief, the antecedents of both reflexives and obligatorily 
controlled PRO must precede what they are coindexed with, and must be 
subjects. If Binding Theory principles apply analogously to both PRO and 
anaphors, then they should apply at the same level of description, in this 
case at LF. The Binding Thoery principles relevant for variables must apply 
also at LF, because the semantic scope of variables is not directly 
indicated at S-Structure (Davison (1984)). 


It is interesting that backward pronominalization is not possible 

except with prior mention of the pronoun referent. In isolation, the 

following sentences contrast in well-formedness: 

i)j]agar raakees .koo Digrii mil gaii, J too woo , waapas calaa 
if R. dat. degree received then 3psg return go 

jaaeegaa "If Rakesh got his degree, he will go back', 
go-fut. (Kachru (1978:13)) 
ii)V*agar us.koo Digrii mil gaii, ] too raakee s. waapas 
if 3psg-DAT degree received then R return 
calaa jaaeegaa. "If he got his degree, Rakesh will go back 


ABBI, Anvita. 1981. Semantic grammar of Hindi. New Delhi: Bahri 

BRESNAN, Joan (ed.). 1982. The mental representation of grammatical 

relations. Cambridge, MA: M.I.T. Press. 
CHOMSKY, Noam 1981. Lectures on government and binding. Dordrecht: Foris 

DAVISON, Alice. 1969. Reflexivization and movement rules in relation to a 

class of Hindi psychological predicates. Chicago Linguistic Society 5. 

. 1981. Syntactic and semantic indeterminacy resolved: a 

mostly pragmatic analysis of the Hindi conjunctive participle. In 

In P. Cole, ed.. Radical pragmatics, pp. 101-28. New York, N.Y.: 


Academic Press. 
. 1984. Syntactic constraints on wh- in situ: constituent 

questions in Hindi-Urdu. Paper read at the igS'l LSA meeting. 
. 1985. Experiencers and patients as subjects in Hindi-Urdu. 

In A. Zide, E. Schiller and D. Magier, eds., Proceedings of the 

conference on participant roles: South Asia and adjacent areas, pp. 

160-78. Bloomington, Ind.: Indiana University Linguistics Club. 
FARMER, Ann. 1984. Modularity in syntax: a study of Japanese and English. 

Cambridge, MA: M.I.T Press. 
GAZDAR, Gerald, Ewan Klein, Geoffrey Pullum and Ivan Sag. 1985. 

Generalized phrase structure grammar. Cambridge, MA: Harvard 

University Press. 
HERMON, Gabriella 1985. Syntactic modularity. Dordrecht: Foris 

Publications . 
KACHRU, Yamuna. 1978. On relative clause formation in Hindi-Urdu. 

Linguistics 207. 5-26. 
. 1981. On the syntax, semantics and pragmatics of the 

conjunctive participle in Hindi-Urdu. Studies in the Linguistic 

Sciences 11.2 pp. 35-50. 
LUTZ, Richard. 1985. The effect of pronoun type on first and second 

language perceptual strategies in Hindi. University of Illinois 

MANZINI, Rita. 1983. Control and control theory. Linguistic Inquiry 14. 

PANDHARIPANDE, Rajeshwari. 1981. Syntax and semantics of the passive in 

selected South Asian languages. University of Illinois dissertation. 
PORIZKA, Vincenc. 1953. Hindstina: Hindi language course. Prague: Statni 

pedagocke nakladatelstvi . 
RAKESH, Mohan. 1962. Antaraal. Delhi: Raajkamal Prakaasan. 
WALLACE, William. 1985. Subjects and subjecthood in Nepali: an analysis of 

Nepali clause structure and its challenges to Relational Grammar and 

Government and Binding. University of Illinois dissertation. 

Studies in the Linguistic Sciences 
Volume 15, Number 2, Fall 1985 


Joseph F. Foster 

University of Cincinnati 


Early attempts to identify special characteristics of 
"primitive" languages failed and linguists' current atti- 
tudes toward the notion imply a definition of primitiveness 
that effectively means that no such attempt can ever succeed. 
The development of markedness and naturalness theories were 
not related to stagial evolution but were nonetheless efforts 
to characterize and distinguish the linguistically normal from 
the possible but rare. That is, they were attempts to try 
again to understand why some languages behave more oddly than 

Some languages have an unusual use of the genitive as 
beneficiary where most use a dative. These seem particularly 
appropriate to the simplex societies they serve, and a full 
understanding of syntactic naturalness may require considera- 
tion of cultural variables normally regarded as extra-lin- 
guistic. is not really possible to understand what is being said 
in linguistics today without knowing the history of the field, 
without knowing what events led up to the current issues that 
are being argued today. — Theodore M. Lightner (1968:46) 

When our scientific ancestors became aware of the range and distri- 
bution of linguistic forms and processes, they found scanty evidence that 
could be adduced in support of stagial evolution. Latin, the language of 
an archaic state level society, had, for instance, five or six cases but 
the Finno-Ugric languages had three times that many and they were spoken 
by peoples still at or recently evolved from the tribal or chiefdom levels 
of sociocultural complexity. Chinese, serving a state level society for 
millenia, had none. The most powerful state level sociocultural system of 
the New World, the Aztec, had a language much like and closely related to 
languages spoken by foraging (hunting/gathering) bands of the Great Basin 
and Plateau. The attempt to relate cultural evolution and levels of socio- 
cultural complexity to grammatical characteristics had by mid-century 
largely been abandoned. The effort to relate cultural evolution to gen- 
erality V. particularity in the lexicon was more tenacious and in 1952 
A. A. Hill felt bound to remind colleagues that "...most modern linguists... 
reject the idea of the inefficiency, formlessness, and overpart icularity 
of primitive speech...." (Hill 1964). He reviewed much of the representa- 
tive literature on the matter and his article of the early fifties provides 
one of the last instances of that period in American linguistics of 
linguists talking to each other about primitiveness. 


The watershed year of 1957 was not far gone, however, before questions 
of evolution in language began to reappear sporadically in the professional 
literature. Dell Hymes returned to the lexical issue in 1961, albeit in a 
somewhat different vein than that in which Hill had left it in 1952, and he 
argued for an evolutionary perspective on lexical expansion and on some 
sociolinguistlc variables. He largely "by-passed" the issue of evolution 
in grammatical characteristics, regarding the denial of such evolution 
however as unproven. Ten years later, Hymes was at least suggesting the 
possibility of evolutionary changes in syntactic processes, for in his 
forward to Swadesh' posthumously published work, he has this to say: 

It is with respect to certain linguistic functions, not all, that 
evolutionary advance can be clearly seen — in the lexical resources 
that differentiate languages of science ,.. .and in the syntactic 
developments that make certain kinds of argument and analysis ex- 
plicitly available. (Hymes in Swadesh 1971:ix) 

The effect of Berlin and Kay's work on cross cultural study of color 
terminology upon linguistics was probably not as great as its effect upon 
cultural anthropology, primarily because of the evolutionary implications 
it contained. (Note especially Berlin 1970 for a succinct statement.) In 
short, they found that societies with few basic color terms tend to be the 
simpler band and tribal level ones. A number of anthropologists regarded 
this work as highly ethnocentric, even though Berlin and Kay also presented 
evidence that color perception and definition is very much alike among all 
humans. That is, the evolutionary implications of their work were not 
with respect to somatic factors but rather with regard to sociocultural 
systems. The resistance seems to have come largely from the fading but 
still noticeable resistance in cultural anthropology to any evolutionary 
theory or explanation of culture whatsoever. This antievolutionary 
position is still evident in British social anthropology and R. M. W. 
Dixon's discussion of cultural differences (Dixon 1980:5-7) seems informed 
by this stance. 

The reappearance of positions favoring an evolutionary perspective 
in the professional literature have dealt largely with the lexicon. With 
respect to syntacticosemantic characteristics, precious little has appeared, 
although some of these scientists have anticipated such investigation. We 
noted Hymes' comment above, and in 1970, Berlin predicted that 

When one looks at language as an adaptive mechanism, there is 
every reason to suspect ... that one will eventually be able to 
make non-obvious and non-trivial generalizations about the 
lexical and grammatical structure of languages spoken by 
primitive peoples which can be interpreted within a cultural 
materialistic theoretical framework. (Berlin 1970:16) 
[Italics mine — JFF] 

But with rare exceptions, including Webb 1977, Foster 1980, and Givon 1979, 
little work toward helping Berlin's prediction come true has been done. 
Most linguists still hold to the view that there are no primitive lan- 
guages, no "survivals", i.e. relics, of primitiveness in otherwise non- 
primitive languages, and that "every language is adequate to Its purpose", 
whatever that might mean. 


To be sure, we sometimes talk of evolution with respect to dlachronic 
subdivision and "speciation" of language families from a protolanguage . 
Goodenough (1980:36-41) is an example. The extinction of a protolanguage 
and its replacement with several related but mutually unintelligible 
daughter languages is somewhat analogous to biological speciation and, 
in that sense, "evolutionary". Unlike biologists and physical anthro- 
pologists, however, linguists make no assumption that the protolanguage 
was in its structure more like the earliest true human languages of the 
paleolithic than are any of the daughter languages. In this respect, 
the use of the term evolution in historical-comparative linguistics is 

This sometime use of evolutionary notions in historical-comparative 
linguistics is also quite different from the notion of cultural evolution 
in cultural anthropology. Foraging bands, such as the I Kung or baMbuti 
Pygmy, have relatively simplex stone and bone technologies, have reciprocal 
exchange and egalitarian access through the group to the basic goods, re- 
sources, and means of production, have little or no political organization 
nor political offices, no warfare, and have religions oriented around 
ritual rather than around morals and ethics. The dominant theoretical 
perspectives, or paradigms, in modern cultural anthropology hold that such 
peoples really are more like the sociocultural systems of paleolithic 
humans than are the complex chiefdoms and even more complex state level 
societies. With this in mind, consider that has often been pointed out, there is today no such thing as 
a 'primitive' language; every language is of approximately equal 
value for the purposes for which it has evolved [italics his — JFF] 
whether it belongs to an advanced or a primitive culture. 
(Samuels 1972:1) 

This statement of course begs the whole question. One might as well 
assert that because lungfish and lemurs are each highly adapted to the 
environments in which they live, there is no such thing as a primitive 
animal. What we ought rather ask is whether there are "purposes" that 
are significantly associated with primitive societies versus complex 
ones, or conversely. The answer of cultural anthropology is of course 
in the affirmative, although many of us would avoid the teleological 
assumptions implied by the choice of the term 'purpose'. Linguistics 
might well ask whether there are linguistic structures primarily associ- 
ated with the one but not the other of these "purposes", or conversely. 
But of course as pointed out at the beginning of this essay, early lin- 
guistics did ask similar questions and with largely negative results. 
That is why most of us would agree with Samuels' claim, although he 
states it in such a way that probably renders it unf alsif iable . 

Since apart from still rather sporadically appearing studies largely 
pertaining to lexica (with exceptions noted above on syntax) linguists 
generally do not believe there are or can be any primitive languages, 
we usually do not write articles for each other about them. To learn 
what linguists think that primitive languages would be like if there 
were any, we must go to the one place where linguists regularly do talk 
about primitiveness-materials for the public and especially for beginning 
students. Roger Brown, dealing primarily with abstraction in the lexicon. 


says that: ". . .pre-llterate languages are not primitive but it is probable 
that the people speaking these languages are nearer the primitive way of 
life than are the members of modern civilized states." (Brown 1968:272). 
Brown cautiously lands on the side of those claiming relative sparsity of 
generic terms in the lexica of primitive peoples and he uses the term prim- 
itive in a way very reminiscent of its use in biology and physical and 
cultural anthropology. Paul Gaeng (1971:8) holds that 

Anthropological research has conclusively shown that even the most 
primitive tribal societies of our time possess highly developed 
structures and complex languages with rich vocabularies. Any dif- 
ferences that may exist between the languages of primitive commu- 
nities and our highly sophisticated cultures lies in the number of 
ideas and concepts that require expression rather than in the way 
in which they are expressed. 

Gaeng's statement has the merit of being empirically falsifiable. Karen Webb 
has given presumption of evidence (1977) that the only languages likely to 
develop transitive verbs of possession like English 'have' are those that 
serve state level societies with differential ownership and stratification 
based on unequal access to the basic means of production. Her sample was 
small but the distribution was statistically significant (p = .0223, Fisher 
Exact Test) by a conservative measure, so that indeed Gaeng's claim may have 
already suffered falsification. 

One of the least dogmatic and most judicious statements about there not 
being any primitiveness in language is that of Ronald Langacker (1973:17). 

Language and culture are closely associated in practice . .but there 
is no reason to believe that any particular type of linguistic 
structure is specially suited to any particular type of culture. 
...In fact... there is no correlation between degree of cultural 
advancement and complexity of linguistic structure.... 

There is a vast difference between asserting that "There are no X." on the 
one hand and reporting on the other that "We have no reason to think there 
are any X." One finds it hard to disagree with the Langacker statement — 
but then one finds it hard to say what exactly the last part of it means. 
The absence of such correlations may be no more than a function of our 
imperfect understanding of language and our inability to agree on any 
measure of "complexity of linguistic structure" — even after a quarter 
century of simplicity metrics. 

The development of simplicity metrics was supposed to help us evaluate 
possible grammars; the development of markedness theory was supposed to 
help us evaluate languages in the sense that it provided a principled and 
systematic way of characterizing unnatural phonological systems and 
accounting for certain kinds of recurring phonological change. It was 
a shock initially to one nurtured as 1 had been in the "all languages are 
equal" school to find us "judging" languages again, but in my mind's ear 
I can yet hear Ted Lightner pointing out of some language with an odd 
rule or very unusual series of segments that "the grammar of this language 


ought to pay in relative complexity for the privilege." He was of course 
absolutely right. A theory of language claiming that Bushman clicks are 
not a possible linguistic consonant series is wrong. But a theory asserting 
them equally as natural to language as /p,t,k./ is bizarre and would now 
be regarded as preciously naive. 

Though it did not incorporate evolutionary considerations, markedness 
theory forced us to deal with the observation that not all phonologies are 
equal. Dell Hymes pointed out that other linguistic inequivalencies exist. 

...differences in adaptation and advance in respect to specific 
criteria do not disappear if ignored. To address the real problems 
of the place of language in social life, one must... deal with 
cases wherein different varieties of language do and mean different 
things. (Hymes in Swadesh 1971 :ix) 

Because, however, of the earlier history of linguistics, most of us have 
had a tendency to assume that any language subject to linguistic analysis, 
i.e. all languages, is a post-primitive language. Thus Greenberg wrote: 

All known human groups have complex languages which exhibit 
essential similarities in their over-all structure. It was 
once thought that peoples with extremely simple technologies, 
so-called primitives, must have languages of a more rudimentary 
type than those of the more technologically advanced peoples. 
This turns out not to hold at all for grammar. ... 

The most important proof of human cultural similarity is 
that languages all exhibit certain basic structural traits, so 
that the techniques of scientific linguistics are applicable 
to all languages. (Greenberg 1977:75) 

This is of course difficult to disagree with, although Webb's possessive 
predication study (Webb 1977) appears to falsify the first paragraph. 
Note that the second paragraph is very like saying that all organisms 
show certain structural and processual traits so that the techniques 
of scientific biology are applicable to all organisms. This is true 
but does not preclude their being significant differences in organisms 
that can be understood in an evolutionary perspective. Cockroaches are 
more primitive than canines, though both are highly "efficient" and 
coherent organisms. We must recall that we abandoned considerations of 
primitiveness when both linguistic theory and cultural anthropology were 
much less sophisticated and powerful than they are today. I do not know 
whether antipassives are more typical of languages spoken in simplex 
societies than complex ones, but if we never ask because by definition 
all describable languages are post-primitive, we'll never know. 

Cambridge University Press' promotional literature on Dixon's 1980 
study of Australian languages says that: "The author stresses that 
these languages are in no sense 'primitive' languages, but have a rich 
and complex grammar with many subtle and distinctive features." For the 
copy writer, linguistic primitiveness means a lack of subtlety and a 
lack of "distinctive features". Now cultures at the band level certainly 
have distinctive features and their peoples are certainly capable of 
"subtlety" by any definition of that word we are likely to agree on. 


Likewise, single-celled organisms have distinctive characteristics, else 
they would all be classed together as one species. In fact, single-celled 
organisms are in a sense "hyper-sophisticated" since they perform with one 
cell functions that in more complex organisms are done with cell specializa- 
tion. If biologists defined their domain of study the way we linguists 
define ours, single-celled organisms would not be regarded as living things. 
If cultural anthropologists operated similarly, most Australian aboriginal 
and other foraging band peoples would not be classed as human cultures I 

Suzette Elgin does not beg the question by defining it out of exist- 
ence — at least not to begin with. 

When contemporary peoples are described as primitive ,.. .usually 
what is meant is that the group described lives in a manner more 
consistent with what we assume to be true of prehistoric man's 
life than with our own.... 

It may be that primitive humans, in the strictest sense of 
the word, spoke primitive languages. But we know nothing at all 
about any such language.... We cannot point to any language and 
say, "This is what a primitive language is (or was) like.... 
(Elgin 1979:202) 

In effect, she says that we don't know what primitive languages might have 
been like and we can't seem to find any extant language for which there is 
a principled basis for classing it as primitive in any scientific (non- 
judgmental, amoral) sense of the word. Now this is largely true and intro- 
ductory texts are hardly the place perhaps to delve into the few instances 
where it may not be true. But then Elgin goes on to close the issue with 
a definition of primitive language which, if we take it seriously, will 
make sure that we never find any. 

For the linguist, then, the term primitive language can have 
only one useful meaning: a language which would be inadequate 
for ordinary human communication. Such a language might have 
no mechanism for adding a new word when a new object was intro- 
duced into the culture of the people speaking it, for example. 

There is probably about the best, most judicious, statement one could come 
up with for a beginning text. But we're not evaluating introductory texts; 
we're using them as a source of information to deduce what our profession 
thinks a primitive language would be like. If we define 'primitive language' 
as only those known not to exist, we have deftly closed the issue without 
having grown any the wiser in the process.'^ 

Since I clearly want to reopen the case, let me make it clear that I 
do not think it will be easy to develop a general set of criteria for 
linguistic primitiveness. It will be hard for several reasons, one being 
our history reviewed in this essay. But there is a difference between 
saying on the one hand that primitive languages don't exist because we 
haven't found any versus saying on the other that they don't exist because 
our definitions of language won't let us find any. One is scientific 
restraint; the other is perilously close to dogma. These definitions are 
themselves the product of a history and cultural process which generated 
a profession that formed its healthy skepticism about primitiveness as a 
result of specific, understandable, conditions. Given the conditions 


under which linguistics grew up, especially in the New World, we could 
hardly have come to any other conclusion. Progress in science requires 
that productive assumptions which have failed of disproof after con- 
siderable effort become something like dogma. As has been quipped before, 
science normally progresses of necessity by closed minds. But it also 
becomes stagnant by closed minds, and the path down which we arrived at 
exclusionary definitions about primitive languages may be quite under- 
standable but it does not compel us to refrain from rethinking the in- 
vestigative strategies reflected in these definitions — particularly since 
we now have much greater sophistication about the nature of grammar, much 
better typological tools, and more powerful insights into the nature of 
cultural evolution than when our history was in the making. 

Elgin's statement and others of its ilk provide a good initial point 
for reopening the investigation. The nonexistence of primitive languages 
is often argued on the basis that every language is "adequate to its 
purpose", for "ordinary human communication", or the like. In adopting 
such a claim, we have left an opening to ask whether some languages are 
in this respect more adequate than others. Is "ordinary human communica- 
tion" the same in industrial states as it is among foraging bandspeople? 
The answer is certainly no, not entirely. We may, that being the case, 
ask whether there be any grammatical processes which seem particularly 
given to association with the expression, predication, or discussion of 
notions and assumptions about the world more commonly found in band and 
egaltarian tribal societies. Conversely, are there any such grammatical 
operations that seem especially given to use with notions or assumptions 
more commonly associated with stratified societies? If the answer to 
either or both these questions turns out affirmative, then there are 
primitive and post-primitive grammatical processes in language, for, 
as Elgin said: "...usually what is meant [by primitive peoples] is that 
the group described lives in a manner more consistent with what we assume 
to be true of prehistoric man's life than with our own." (Elgin 1979:202) 

The work of Webb (1977) with regard to intransitive v. transitive 
predication of possession has been noted above. Recall that there is a 
statistically significant tendency for languages with transitive predica- 
tion as the usual way of saying 'I have a new house.' to be spoken by 
state level societies. I have proposed elsewhere (Foster 1980) that a 
structure which predicates possession with possessor as subject of a 
transitive verb in the nominative case is a significantly rather than 
superficially different kind of structure from the predication of 
possession by a transitive ergative structure with possessor "subject" 
in the ergative case. This being so, then transitive active predication 
of possession on the one hand as contrasted with either transitive 
ergative or intransitive predication of possession on the other fit 
Hymes' "...cases wherein different varieties of language do and mean 
different things". From an entirely different set of considerations, 
Talmy Givon has proposed that SOV languages were more apt to have served 
our prehistoric lithic ancestors than other types of basic word order 
languages. (Givon 1979. See especially Chapters 7 and 8.) As a final 
example, there are at least two languages, De ' kwana of Venezuela and 
Dyirbal of Australia, which use the genitive case as the normal case 
of the recipient-beneficiary where most languages use a dative and 


despite the fact that both these languages do have a dative case. Consider 
for instance the English utterance 

(1) @* The man gave the woman's beans. 

The symbol '(3' represents a sentence which, whatever its grammaticality , 

is culturally anomalous. As (1) stands out of any context, it is not an 

English sentence. If it means anything at all, it means that the man took 

beans properly belonging to the woman and gave them to some third party. 

In a dialogue like (2), it works linguistically, however illegal or acustomary 

the implied activity might be. 

(2) "I'm giving my squash to the church bazaar and Sally's 

giving her tomatoes. What 're you gonna give?" 

@ "Oh, I'm giving my neighbor's beansi" 

In no instance would English (1) ever be taken to mean that the recipient 
of bean-giving was the woman possessor of the beans, marked with the gen- 
itive case suffix. 

Similarly in rural Welsh, sentence (3) is anomalous and woman, the 
possessor of the rabbit, is not understood to be the recipient. 

(3) Welsh: (?* Mi rodd ef y gwr y geinach y ddynes. 

Past give he the man the rabbit of: the woman 
'The man gave the woman's rabbit.' 

In German there is in principle a possible ambiguity since in the feminine 
declension, the case for genitive and for dative are homophonous, although 
the ambiguity chance is somewhat reduced by the normal tendency of the 
genitive to follow the possessed. 

(4) German: Der Mann hat der Frau Bohnen gegeben. 

the man has DAT/GEN: the woman beans given 

a. 'The man gave the woman beans.' 

b. 'The man gave the woman's beans.' 

In fact, all native speakers of German take the (a) interpretation, and 
no speaker likes the sentence 

(5) German: *Der Man hat die Bohnen der Frau gegeben. 

if it means that the woman is both the possessor and recipient of the beans. 

Leaving Indoeuropean, we find Turkish with both a genitive and an allo- 
dative case and a similar situation. 

(6) Turkish: @Adam hanum - un tavjan - wn - «■ verdi. 

man woman - GEN rabbit - POSSESSED - OBJ gave 
'The man gave the woman's rabbit.' 

In general, a recipient is in the dative case and not the genitive. Now, 
Dyirbal, like English, Welsh, German, and Turkish, makes a dative/genitive 


distinction and Dyirbal, like Turkish, has an allative-dative case. However, 
in the Dyirbal equivalent, the recipient and possessor is in the genitive. 
(Dyirbal examples are all from Dixon 1972.)-^ That is, the woman is the 
owner and recipient of the beans, and the sentence is perfectly normal and 
the usual way of saying this in Dyirbal. 

(7) Dyirbal: ^ 

balam miran banun ^ugumbil - ^u bangul yafa - i}gu waga-n. 
CLASS beans :NOM CLASS woman-GEN CLASS man - ERG gave /gives 
'The man gave the woman's beans.' 

The English translation is of course our anomalous example (1) . Perhaps a 
closer translation would be 

(8) The man gave the woman's beans to her. 

Dyirbal, however, does not say it that way, primarily because the Dyirbal 
and European make quite different assumptions about what giving is, and 
the normal Dyirbal sentence does not imply, in contradistinction to the 
English in example (8) , that the beans would have been given by a rational 
person to anyone else, nor that the option not to give the beans exists. 

The key to the linguistically unusual use by Dyirbal of the genitive 
in these instances is that in Dyirbal, the simple genitive is used to 
indicate the possessor by right . Dixon refers to this structure as 
proleptic , and indeed to us it does involve a kind of treating of a 
future event as though it had already happened. But in a less ethno- 
centric sense, it is not really quite the same thing as our philosophical 
notion of prolepsis. Primitive societies are sacred societies in which 
everything is in micro- or macrocosm really the same thing. The giving 
here is not in the Western sense at all but a transfer of material to 
the control of the person who is the rightful controller of it. This 
right is determined, not through a market economy, not through that 
person's own effort or agency, but rather through that person's relation 
to the bestower of control, the "giver" as defined by the kinship system. 
Dixon points out (1972:30-1, 237) that spontaneous and "unnecessary" 
giving is uncommon among the Dyirbal but that socially necessary giving 
in the sense of control transfer goes on all the time. For instance, a 
relative of a hunter has a right to a portion of the meat that hunter 
brings in and thus that meat already belongs to that relative. Possession 
among the Dyirbal of comestibles and other material is overwhelmingly 
an ascribed relationship, not an acquired one by an individual's own 
efforts as it is in state level societies. Dixon in fact noted (1973) 
this cultural difference with respect to non-nuclear verbs of giving 
(like 'carry') but has not related it to an evolutionary pattern. 
Indeed, if his comments in his 1980 (4-7) work are an indication, he 
would probably reject any notion of cultural evolution whatsoever, let 
alone linguistic evolution. Bu